Differential Motion Between Mediastinal Lymph Nodes and Primary Tumor in Radically Irradiated Lung Cancer Patients

International Journal of

Radiation Oncology biology

physics

www.redjournal.org

Physics Contribution

Differential Motion Between Mediastinal Lymph Nodes and Primary Tumor in Radically Irradiated Lung Cancer Patients Eva E. Schaake, MD, PhD,*,y Maddalena M.G. Rossi, MSc,y Wieneke A. Buikhuisen, MD,* Jacobus A. Burgers, MD, PhD,* Adrianus A.J. Smit, MD,z Jose´ S.A. Belderbos, MD, PhD,y and Jan-Jakob Sonke, PhDy Departments of *Thoracic Oncology and yRadiation Oncology, The Netherlands Cancer Institute; and z Department of Pulmonary Disease, Onze Lieve Vrouwe Gasthuis, Amsterdam, The Netherlands Received Mar 17, 2014, and in revised form Jul 16, 2014. Accepted for publication Jul 25, 2014.

Summary Detailed information about lymph node motion variability and differential motion in relation to the primary tumor is lacking in patients with locally advanced lung cancer, limiting accurate margin design. With the use of fiducial markers and daily 4-dimensional cone beam computed tomography in 51 patients, detailed analysis revealed considerable interfractional differential motion, modest intrafractional variability, and stable respiratory motion requiring margins up to 1.4 cm. A correction strategy based on

Purpose/Objective: In patients with locally advanced lung cancer, planning target volume margins for mediastinal lymph nodes and tumor after a correction protocol based on bony anatomy registration typically range from 1 to 1.5 cm. Detailed information about lymph node motion variability and differential motion with the primary tumor, however, is lacking from large series. In this study, lymph node and tumor position variability were analyzed in detail and correlated to the main carina to evaluate possible margin reduction. Methods and Materials: Small gold fiducial markers (0.35  5 mm) were placed in the mediastinal lymph nodes of 51 patients with non-small cell lung cancer during routine diagnostic esophageal or bronchial endoscopic ultrasonography. Fourdimensional (4D) planning computed tomographic (CT) and daily 4D cone beam (CB) CT scans were acquired before and during radical radiation therapy (66 Gy in 24 fractions). Each CBCT was registered in 3-dimensions (bony anatomy) and 4D (tumor, marker, and carina) to the planning CT scan. Subsequently, systematic and random residual misalignments of the time-averaged lymph node and tumor position relative to the bony anatomy and carina were determined. Additionally, tumor and lymph node respiratory amplitude variability was quantified. Finally, required margins were quantified by use of a recipe for dual targets. Results: Relative to the bony anatomy, systematic and random errors ranged from 0.16 to 0.32 cm for the markers and from 0.15 to 0.33 cm for the tumor, but despite similar ranges there was limited correlation (0.17-0.71) owing to differential motion.

Reprint requests to: Jan-Jakob Sonke, PhD, Department of Radiation Oncology, The Netherlands Cancer Institute, Plesmanlaan 121, 1066 CX Amsterdam, The Netherlands. Tel: þ31-20-512 1731; E-mail: [email protected] Conflict of interest: The Department of Radiation Oncology of the Netherlands Cancer Institute receives research grants and software license Int J Radiation Oncol Biol Phys, Vol. 90, No. 4, pp. 959e966, 2014 0360-3016/$ - see front matter Ó 2014 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.ijrobp.2014.07.038

fees from Elekta Oncology Systems Ltd. The authors report no other conflict of interest. Supplementary material for this article can be found at www.redjournal.org.

960

International Journal of Radiation Oncology  Biology  Physics

Schaake et al.

carina registration allows margin reduction up to 27%.

A large variability in lymph node amplitude between patients was observed, with an average motion of 0.56 cm in the cranial-caudal direction. Margins could be reduced by 10% (left-right), 27% (cranial-caudal), and 10% (anteroposterior) for the lymph nodes and 2%, 15%, and 7% for the tumor if an online carina registration protocol replaced a protocol based on bony anatomy registration. Conclusions: Detailed analysis revealed considerable lymph node position variability, differential motion, and respiratory motion. Planning target volume margins can be reduced up to 27% in lung cancer patients when the carina registration replaces bony anatomy registration. Ó 2014 Elsevier Inc.

Introduction Metastatic invasion to the mediastinal lymph nodes strongly correlates with the life expectancy of lung cancer patients (1). When mediastinal lymph nodes are involved, the treatment of choice consists of concurrent chemoradiation of the primary lung tumor and affected mediastinal lymph nodes. For optimal irradiation, it is necessary to adequately account for geometric uncertainties such as baseline shifts, differential, and respiratory motion (2). Traditionally, generous safety margins are applied to account for such geometric uncertainties in the absence of adequate correction strategies exposing nearby organs at risk to high radiation doses resulting in considerable toxicity (3-7), such as esophagitis and pneumonitis. Lymph nodes are soft tissues surrounded by vessels and other soft mediastinal tissues, and they are therefore difficult to localize on in-room imaging series such as cone beam computed tomographic (CBCT) scans. It is therefore unclear whether current planning target volume (PTV) margins of involved lymph nodes sufficiently account for their variability. Current margins are based on tumor position variability only. In a previous proof of principle analysis of a first patient group of 14 patients, we showed that it is feasible and easy to place gold fiducial markers in mediastinal lymph nodes during routine diagnostic endoscopic ultrasound procedures (8). Subsequently, the cohort was extended to 51 patients to improve statistical power, quantify differential motion between lymph nodes and primary tumor, evaluate intrafractional variation and registration accuracy, and finally properly advise on margins. The lymph nodes were mapped according to Naruke et al (9) with different stations. The International Association for the Study of Lung Cancer (IASLC) adopted this into a lymph node map (10). The aim of this study was to perform a detailed analysis of lymph node and primary tumor position variability, differential motion, and required safety margins.

Methods and Materials A single-center prospective cohort study, opened in 2010, was performed in patients with non-small cell lung cancer (NSCLC) who were planned for radical radiation therapy to determine mediastinal lymph node position variability by the

use of gold fiducial markers (0.35 mm  5.0 mm [RadioMed, Barlett, TN] from now on referred to as marker). Patients who were medically fit (World Health Organization 0-2), were 18 years or older, had not previously undergone mediastinal surgery or chest irradiation, and underwent endobronchial ultrasound (EBUS) or transesophageal endoscopic ultrasound (EUS) guided mediastinal lymph node aspiration (11) for diagnostic purposes and were scheduled for radical radiation therapy or chemoradiation as primary treatment were eligible for inclusion in this study. Lymph nodes were selected for marker implantation on the basis of radiologic suspicion of tumor invasion and accessibility. Written informed consent was obtained from all patients according to International Conference of Harmonisation/Good Clinical Practice (ICH/ GCP) and national and local regulations. This study was approved by the institute’s medical ethics committee. The primary endpoint of the study was to quantify lymph node position and motion variability over the course of radiation therapy in NSCLC patients.

Radiation therapy preparation Every patient received radical intensity modulated radiation therapy to a dose of 66 Gy in 24 fractions with an overall treatment time of 32 days. Patients were stabilized by an armrest (in-house modified armrest and knee support, Civco, Orange City, IA). The treatment schedule could be adjusted on the basis of inclusion in clinical trials, mean lung dose, or the patient’s comorbidity. A 4-dimensional (4D) planning computed tomography (CT) scan with intravenous contrast medium (0.3-cm slice thickness) was acquired, and the midposition CT scan was subsequently reconstructed (12) for delineation and dose calculation. The planned dose distribution encompassed the primary tumor and lymph nodes that were pathologic or suspected positive by 18fluorodeoxyglucose positron emission tomography. The clinical target volume (CTV)-to-PTV margin was 12 mm for the lymph nodes, and for the primary tumor the CTV-to-PTV margin was 12 mm þ 0.25 A, with A being the peak-to-peak amplitude of the gross tumor volume (full range of motion: minimum to maximum) derived from the 4D planning CT. Daily 4D CBCT scans were acquired, and bony anatomy registration was used for online setup error correction. Once a week, a CBCT scan was performed immediately after treatment delivery to quantify the intrafractional motion.

Volume 90  Number 4  2014

Lymph node and tumor position variability analysis Lymph node marker registration was described previously in a feasibility study that demonstrated negligible marker migration (8). In short, for each phase of a 4D CBCT scan, the position of the marker was local-rigidly registered (translations only) to the planning CT, and the time-weighted mean lymph node position was calculated and corrected for bony anatomy misalignments. Additionally, the peak-to-peak breathing amplitude was calculated by its excursion over the phases. Interfractional variability was quantified in terms of grand mean (GM), systematic errors (S), and random errors (s) (13). The combined effect of intrafractional variability and registration inaccuracy was assessed by the difference between pretreatment and posttreatment CBCT scan registrations corrected for the prescribed couch shift. In a similar way, the main carina and tumor displacements were quantified by a local rigid 4D gray value registration (translations only). All registrations were visually verified, and occasionally the starting point, the shaped region of interest, the registration result, or a combination of these were manually adjusted to obtain a clinically meaningful result in the presence of tumor regression. More specifically, if a regressing tumor was attached to an adjacent structure, this structure was included in the region of interest with the aim of aligning this side of the lesion. If the regressing tumor was surrounded by lung tissue, the aim was to align the center of the residual mass with the center of the mass of the gross tumor volume in the planning CT (Appendix, available at www.redjournal.org). Analysis of the registration accuracy by use of a 4D-adapted full-circle method is described in the Appendix. Additionally, a carina-based correction strategy (CbCS) and tumor-based correction strategies (TbCS) were simulated, and residual marker and tumor position variabilities were quantified and compared with the bony anatomyebased correction strategy (BAbCS). Pearson product-moment correlation coefficients between the patient-specific systematic marker and the tumor displacement were analyzed for both BAbCS and CbCS. Analysis was performed in Matlab-7 (MathWorks Inc, Natick, MA). Differences in grand mean, systematic and random variations were tested by Student t test, c2 test, and sign rank test, respectively.

Margins Traditionally, PTV margins are calculated for a single target. In this study, 2 separate targets were analyzed. The derived PTV margin M to ensure that the lymph nodes and the primary tumor simultaneously receive at least 95% of the prescribed dose for 90% of patients was calculated as:
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi  h i s2 þ s2p  sp MZ2:79S þ 1:64 The factor 2.79 instead of the more familiar 2.5 was taken from Table 2 in the article by van Herk et al (14) for a 3-dimensional error distribution and a

Differential motion in NSCLC patients

961

pffiffiffiffiffiffiffi confidence level of 0:9z0:95, conservatively assuming uncorrelated uncertainties between lymph nodes and primary tumor. The overall systematic and random errors were calculated separately for lymph nodes and qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi primary tumor as SZ ½S2B þ S2TD þ S2L þ S2I  and pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi sZ ½s2B þ s2L þ s2I þ s2R , where the subscript B refers to baseline variation, TD refers to target definition uncertainty (15), L refers to localization accuracy, I refers to intrafractional variability, and R refers to respiratory motion. In this study, sR will be approximated by A/3, with A the peak-to-peak amplitude. For the lymph nodes sp Z .32 cm (16) and for the primary tumor sp Z .64 cm were used to describe the width of the penumbra (17).

Results Between September 2010 and August 2013, markers were placed in 63 patients. Fifty-four of these patients actually Table 1

Characteristics of patients analyzed All patients (nZ51) Characteristic

Age, y Sex Male Female Stage IA IB IIA IIB IIIA IIIB IV Histology/cytology Adenocarcinoma Squamous cell carcinoma Undifferentiated NSCLC Radiation dose No. of fractions Fraction dose Marker in lymph node stations, n II Left IV Left IV Right V VII VIII XI Chemotherapy None Concurrent daily low-dose cisplatin Other

n or median

% or range

64

36-85

27 24

53 47

0 1 3 2 24 20 1

0 2 6 3 46 38 2

15 19 17

30 37 33

24 2.75

17-30 2-3.4

6 11 10 2 17 2 3

12 22 20 4 33 4 6

5 44

10 86

2

4

Abbreviation: NSCLC Z non-small cell lung cancer.

962

Schaake et al.

received radical irradiation. Three of these patients could not be analyzed because of inconsistencies in the protocol logistics and were excluded. The EUS and EBUS guided marker placement occurred without complications, and no markers were lost during placement or treatment. The patient and tumor characteristics of the 51 patients analyzed are listed in Table 1. All inserted markers were well visible on planning CT and CBCT (Fig. 1). An average of 22 CBCT scans was acquired per patient (range, 17-24). Intrafractional motion was measured in 49 patients, with an average of 4 posttreatment CBCT scans per patient (range, 3-6).

International Journal of Radiation Oncology  Biology  Physics

left-right (LR) and anteroposterior (AP) directions but statistically significant for the CC direction: 0.08 cm (13% smaller during treatment; PZ.04) for the markers and 0.13 cm (24% smaller during treatment; PZ.002) for the tumor. The systematic amplitude differences ranged from 0.19 to 0.27 cm for the markers and 0.12 to 0.28 cm for the tumor. Random variations were typically 40% to 53% smaller. Registration accuracy was analyzed and was below 1 mm (standard deviation) for all structures (Appendix, available at www.redjournal.org). It therefore had a limited impact on these results.

Carina-based and tumor-based corrections Lymph node and primary tumor variation Grand mean baseline variation was statistically insignificant and 0.05 cm in all directions for both lymph nodes and tumor (Table 2). The systematic baseline variation ranged from 0.20 to 0.32 cm for the markers and 0.15 to 0.33 cm for the tumor. Random errors were up to 25% smaller than systematic errors. The mean peak-to-peak amplitude of the lymph nodes during treatment was 0.56 cm in the cranial-caudal (CC) direction and on the population level very similar to that of the tumor: 0.55 cm (Table 3). Grand mean differences relative to the 4D planning CT scan were negligible for the

Carina-based corrections successfully reduced the interfractional position variability of both lymph nodes and tumors compared with bony anatomyebased corrections (Table 2). For the markers, these reductions (or increases for negative numbers) were 22% (LR; PZ.09), 59% (CC; P<1e4), and 24% (AP; PZ.06) systematically and 33% (P<1e4), 40% (P<1e4), and 24% (PZ.002) randomly. For the tumor this was 3.9% (PZ.79), 27% (PZ.03), and 13% (PZ.35) systematically and 9.7% (PZ.88), 19% (PZ.0002), and 16% (PZ.005) randomly. Note that a carina-based correction induced variability of the bony anatomy that otherwise would be nearly perfectly

Fig. 1. Color fusion of a midposition planning computed tomography scan (purple) and single phase of a 4-dimensional cone beam computed tomography scan (green) illustrating a baseline shift of both lymph node (green purple marker displacement visible in coronal and sagittal view, top and bottom right) and primary tumor (large green purple tumor displacement visible in coronal view, top left).

Volume 90  Number 4  2014

Differential motion in NSCLC patients

Table 2 Grand mean, systematic, and random interfractional position variability relative to bony anatomy, carina, and tumor Variability

LR (cm)

CC (cm)

AP (cm)

Residual lymph nodes and tumor position variability after BAbCS Marker GM 0.023 0.040 0.048 S 0.20 0.32 0.21 s 0.19 0.24 0.16 Tumor GM 0.004 0.049 0.031 S 0.15 0.33 0.23 s 0.16 0.26 0.25 Residual bony anatomy, lymph nodes, and tumor position variability after CbCS Bone GM 0.041 0.033 0.001 S 0.12 0.28 0.17 s 0.15 0.21 0.14 Marker GM 0.018 0.007 0.048 S 0.15 0.13 0.16 s 0.13 0.14 0.12 Tumor GM 0.036 0.016 0.031 S 0.16 0.24 0.20 s 0.18 0.21 0.21 Residual bony anatomy and lymph nodes position variability after TbCS Bone GM 0.005 0.049 0.031 S 0.15 0.33 0.23 s 0.16 0.26 0.25 Marker GM 0.019 0.009 0.017 S 0.23 0.25 0.21 s 0.22 0.23 0.21 Abbreviations: AP Z anteroposterior; BAbCS Z bony anatomyebased correction strategy; CbCS Z carina-based correction strategy; CC Z cranial-caudal; GM Z grand mean; LR Z left-right; s Z random; S Z systematic; TbCS Z tumor-based correction strategy.

positioned, with systematic and random variability ranging from 0.12 to 0.28 cm. For a tumor-based correction, the positional variability of the lymph nodes reduced by 16% (LR; PZ.31), 22% (CC; PZ.09), and 2% (AP; PZ.89) systematically and 15% (PZ.05), 1% (PZ.56), and 31% (PZ.04) randomly, compared with a bony anatomyebased correction (Table 2).

Intrafraction variation Intrafraction variability was considerably smaller than interfraction variability (Table 4). Nevertheless, the submillimeter grand mean values were statistically significant for the CC and AP directions for both the markers and the tumor, indicating a small drift to the cranial and posterior

963

Table 3 Interfractional peak-to-peak amplitude statistics for the population and grand mean (GM), systematic (S), and random (s) amplitude changes relative to 4-dimensional planning computed tomography for both marker and primary tumor Amplitude statistics Marker Population mean Population minimum Population maximum GM S

s

Tumor Population mean Population minimum Population maximum GM S

s

LR (cm) 0.23 0.10 0.50 0.002 0.19 0.1 0.14 0.03 0.35 0.02 0.12 0.07

CC (cm)

AP (cm)

0.56 0.12 1.7 0.08 0.27 0.13

0.24 0.10 0.53 0.01 0.19 0.11

0.55 0.04 2.26 0.13 0.28 0.13

0.21 0.06 0.74 0.02 0.14 0.08

Abbreviations: AP Z anteroposterior; CC Z cranial-caudal; LR Z left-right.

side. Systematic and random intrafraction variations ranged from 0.15 to 0.21 cm and were 20% to 45% smaller than interfractionally. Intrafraction amplitude variation was also quite small, with grand mean changes <0.05 cm and systematic variation <0.1 cm. Random amplitude variations were somewhat larger, 0.17 cm, but were likely affected by the registration accuracy of 0.15 cm as described in the Appendix (available at www.redjournal.org).

Correlation between marker and tumor The correlation between the marker and tumor systematic position variability was determined in the 3 orthogonal directions. For bony anatomyebased corrections, these correlations were 0.17 (LR; PZ.23), 0.71 (CC; P<1e4), and 0.55 (AP; PZ<1e4). For carina-based corrections, the correlations were 0.06 (LR; PZ.69), 0.21 (CC; PZ.16), and 0.34 (AP; PZ.016).

Margins The required PTV margins depend on the accuracy of delineation, the patient’s respiratory motion, and the correction strategy. In Figure 2, the required PTV margins are shown for both the lymph nodes and tumor. Margins were plotted for varying respiratory amplitude because this varies across patients and for varying delineation uncertainties because these depend on tumor location, size, and imaging modality (18). For the population average respiratory amplitudes and an assumed target definition uncertainty STD Z 0.2 cm (18) (valid for parts of the tumor with a clear boundary), the margins for a BAbCS were

964

International Journal of Radiation Oncology  Biology  Physics

Schaake et al.

Table 4 Grand mean, systematic, and random intrafractional position and amplitude variability of lymph nodes and tumor Variability Position variability Marker GM S

s

Tumor GM S

s

LR (cm)

CC (cm)

AP (cm)

0.012 0.19 0.17

0.085 0.17 0.21

0.057 0.15 0.15

0.019 0.18 0.16

0.067 0.19 0.21

0.07 0.16 0.16

Amplitude variability Marker GM 0.005 S 0.07 s 0.13 Tumor GM 0.017 S 0.03 s 0.05

0.043 0.10 0.17

0.012 0.08 0.14

0.046 0.09 0.13

0.012 0.04 0.07

Abbreviations: AP Z anteroposterior; CC Z cranial-caudal; GM Z grand mean; LR Z left-right; s Z random; S Z systematic.

1.1 cm (LR), 1.4 cm (CC), and 1.0 cm (AP) for lymph nodes and 0.9 cm (LR), 1.4 cm (CC), and 1.1 cm (AP) for tumor. For a CbCS, margin reductions of 10%, 27%, and 10% for the lymph nodes and 2%, 15%, and 7% for the tumor are achievable. For a TbCS, no margin reduction is achievable for the lymph nodes, but for the tumor, reductions of 16%, 37%, and 30% are achievable.

Discussion Lymph node and tumor position variability and differential motion were analyzed in detail. By use of a margin recipe for multiple targets, a CbCS was shown to be superior to a BAbCS. Margins can be decreased up to 27% in the CC direction by using the main carina instead of the bony anatomy as a surrogate for the tumor and lymph nodes. In recent years, a few studies have been published that quantify lymph node position variability. The interfraction variability of lymph nodes found in this study was up to 1 mm different from that in our proof of principle study, which included only 14 analyzed patients (8). This could be expected with the low number of included patients in the pilot study. Similarly, Jan et al (19), Roman et al (20), and Weiss et al (21) published analyses on small patient cohorts based on delineated structures. weekly scans, or both and reported similar (20) or larger variability than found in this study (19, 21), likely because of delineation uncertainty, limited statistical power, or both. In patients without large anatomic changes, however, Weiss et al (21) found somewhat smaller variability. The baseline shifts quantified for the primary tumor are within 0.1 cm of previously published results (2, 17).

The systematic amplitude variability was larger than the random variability, illustrating the imperfection of the 4DCTebased amplitude assessment during treatment preparation, which is based on a single respiratory cycle, and is thus more susceptible to breathing irregularities than a 4DCBCT-based amplitude assessment which is based on dozens of cycles. Also, there was a small reduction of the average amplitude. The 4D-CT was obtained with use of a respiratory sensor (a thermocouple inserted into the entry of a regular oxygen mask), which possibly influences the respiratory behavior while being absent during 4D-CBCT acquisition. The low random amplitude variability of both tumor and lymph node amplitude illustrates the day-to-day stability of the patient’s respiratory behavior, which is in line with observations by Sonke et al (17) and Rit et al (22). In this study, intrafraction variability was measured weekly by a posttreatment scan. The variability assessed by this posttreatment scan, therefore, is the combined effect of imperfect corrections (localization variability) and position changes during treatment delivery (true intrafraction motion). This might explain that the reported intrafractional position variability is somewhat larger than the intrafractional variability reported for stereotactic body radiation therapy (23), despite the shorter delivery time for conventionally fractionated radiation therapy. The intrafractional amplitude variability was generally quite small, but the random intrafractional amplitude variability of the lymph nodes was larger than the interfractional variability. Note, however, that the intrafractional variation was the result of the difference in amplitude between the precorrection and posttreatment scans and thus more susceptible to measurement uncertainties as described in the Appendix (available at www.redjournal.org). The CbCS allows considerable margin reduction compared with a BAbCS and was most effective in the CC direction, both for the lymph nodes and for the primary tumor. A TbCS strategy did not achieve a significant reduction of lymph node position variability despite a reasonable correlation, and therefore margin calculations were omitted. Such a correction strategy, however, reduces the required margin for the primary tumor considerably (2, 17). The impact of achievable margin reduction for the various targets and the corresponding optimal correction strategy is likely patient specific, depending on the size and location of the primary tumor, number of involved lymph nodes, exposure to organs at risk, and the patient’s health and could thus be the subject of further study. The calculated margin (Fig. 2) increased more rapidly with respiratory amplitude for the lymph nodes than for the tumor. This was a result of the sharper penumbra (sp) modeled for the lymph nodes. It is assumed here that the lymph nodes are surrounded by water-equivalent material, whereas the tumor is surrounded by lung. If the tumor is adjacent to the mediastinum or thorax wall, somewhat larger margins are required. This study had some limitations. A gold marker was used as a surrogate of the lymph node position, yet markers

Volume 90  Number 4  2014

Differential motion in NSCLC patients

965

Fig. 2. Planning target volume (PTV) margin for the mediastinal lymph nodes (A, C) and primary lung tumor (B, D) as a function of the peak-to-peak amplitude (A, B) (at target dimension uncertainty Z 0.2 cm) or target delineation variation (C, D) (at population average respiratory amplitudes). Solid lines represent margins corresponding to bony anatomyebased correction strategy. Dashed lines represent margins corresponding to carina-based correction strategy. AP Z anteroposterior; CC Z cranial-caudal; LR Z left-right. are not lymph nodes. Therefore, rotations could not be assessed but have a limited impact of 0.026 cm/degree for a 3-cm lesion. Similarly, marker migration and anatomic changes have limited impact on the precision of lymph node motion detection (8). Therefore, implanted markers are likely more representative than a delineation of the very-low-contrast structure in 4D CBCT scans. The margin recipe for dual targets used in this study was based on the assumption of uncorrelated motion of tumor and lymph nodes, whereas correlations up to 0.71 were observed. However, even for such a correlation, the factor for the systematic errors S exceeded 2.7 such that this assumption remains reasonable. Only a limited amount of lymph node stations were reached by EUS and EBUS. We have therefore not been able to distinguish the differences in position variability for different lymph node stations. Some patients receiving a marker didn’t received radical irradiation and were thus not analyzed. One patient received radical resection after EUS revealed diseasenegative lymph nodes, and 7 patients did not receive radical irradiation because of metastatic disease discovered during work-up. Daily online CBCT became our standard practice in 2012 after our pilot study indicated that our lymph node margins were inadequate for an offline correction protocol.

Besides more accurate positioning, such an online protocol allows close monitoring of anatomic changes such as increased or decreased atelectasis and pleural effusion and (associated) displacements of primary tumor, lymph nodes, or both. Such changes over the course of treatment can be resolved with adaptive radiation therapy. On the basis of this study, a CbCS has been clinically implemented. PTV margins, however, have not yet been changed. Note that a CbCS increases the position variability of the spinal cord compared with an online BAbCS. On the other hand, the resulting position variability is quite similar to those reported for an offline BAbCS and thus similar planning tolerance limits may be applied. We have shown that a CbCS allows reduction of margins and thereby possibly reduces toxicity in the mediastinum, without the expectation of a decrease in local control. This hypothesis should, however, be tested in a clinical trial because it may affect control of microscopic disease.

References 1. Dehing-Oberije C, De RD, van der Weide H, et al. Tumor volume combined with number of positive lymph node stations is a more important prognostic factor than TNM stage for survival of non-smallcell lung cancer patients treated with (chemo)radiotherapy. Int J Radiat Oncol Biol Phys 2008;70:1039-1044.

966

Schaake et al.

2. Sonke JJ, Lebesque J, van Herk M. Variability of four-dimensional computed tomography patient models. Int J Radiat Oncol Biol Phys 2008;70:590-598. 3. Graham MV, Purdy JA, Emami B, et al. Clinical dose-volume histogram analysis for pneumonitis after 3D treatment for non-small cell lung cancer (NSCLC). Int J Radiat Oncol Biol Phys 1999;45:323-329. 4. Jenkins P, D’Amico K, Benstead K, et al. Radiation pneumonitis following treatment of non-small-cell lung cancer with continuous hyperfractionated accelerated radiotherapy (CHART). Int J Radiat Oncol Biol Phys 2003;56:360-366. 5. Rancati T, Ceresoli GL, Gagliardi G, et al. Factors predicting radiation pneumonitis in lung cancer patients: A retrospective study. Radiother Oncol 2003;67:275-283. 6. Seppenwoolde Y, Lebesque JV, de Jaeger K, et al. Comparing different NTCP models that predict the incidence of radiation pneumonitis: Normal tissue complication probability. Int J Radiat Oncol Biol Phys 2003;55:724-735. 7. Tsujino K, Hirota S, Endo M, et al. Predictive value of dose-volume histogram parameters for predicting radiation pneumonitis after concurrent chemoradiation for lung cancer. Int J Radiat Oncol Biol Phys 2003;55:110-115. 8. Schaake EE, Belderbos JS, Buikhuisen WA, et al. Mediastinal lymph node position variability in non-small cell lung cancer patients treated with radical irradiation. Radiother Oncol 2012;105:150-154. 9. Naruke T, Suemasu K, Ishikawa S. Lymph node mapping and curability at various levels of metastasis in resected lung cancer. J Thorac Cardiovasc Surg 1978;76:832-839. 10. Rusch VW, Asamura H, Watanabe H, et al. The IASLC lung cancer staging project: A proposal for a new international lymph node map in the forthcoming seventh edition of the TNM classification for lung cancer. J Thorac Oncol 2009;4:568-577. 11. Lennon AM, Penman ID. Endoscopic ultrasound in cancer staging. Br Med Bull 2007;84:81-98. 12. Wolthaus JW, Schneider C, Sonke JJ, et al. Mid-ventilation CT scan construction from four-dimensional respiration-correlated CT scans for radiotherapy planning of lung cancer patients. Int J Radiat Oncol Biol Phys 2006;65:1560-1571.

International Journal of Radiation Oncology  Biology  Physics 13. van Herk M. Errors and margins in radiotherapy. Semin Radiat Oncol 2004;14:52-64. 14. van Herk M, Remeijer P, Rasch C, et al. The probability of correct target dosage: Dose-population histograms for deriving treatment margins in radiotherapy. Int J Radiat Oncol Biol Phys 2000;47: 1121-1135. 15. McKinzie A. Defining the PTV and PRV: New ideas about old problems [abstract]. Radiother Oncol 2004;73(Suppl1):S203. 16. Witte MG, van der Geer J, Schneider C, et al. The effects of target size and tissue density on the minimum margin required for random errors. Med Phys 2004;31:3068-3079. 17. Sonke JJ, Rossi M, Wolthaus J, et al. Frameless stereotactic body radiotherapy for lung cancer using four-dimensional cone beam CT guidance. Int J Radiat Oncol Biol Phys 2009;74:567-574. 18. Steenbakkers RJ, Duppen JC, Fitton I, et al. Reduction of observer variation using matched CT-PET for lung cancer delineation: A threedimensional analysis. Int J Radiat Oncol Biol Phys 2006;64:435-448. 19. Jan N, Balik S, Hugo GD, et al. Interfraction displacement of primary tumor and involved lymph nodes relative to anatomic landmarks in image guided radiation therapy of locally advanced lung cancer. Int J Radiat Oncol Biol Phys 2014;88:210-215. 20. Roman NO, Shepherd W, Mukhopadhyay N, et al. Interfractional positional variability of fiducial markers and primary tumors in locally advanced non-small-cell lung cancer during audiovisual biofeedback radiotherapy. Int J Radiat Oncol Biol Phys 2012;83: 1566-1572. 21. Weiss E, Robertson SP, Mukhopadhyay N, et al. Tumor, lymph node, and lymph node-to-tumor displacements over a radiotherapy series: Analysis of interfraction and intrafraction variations using active breathing control (ABC) in lung cancer. Int J Radiat Oncol Biol Phys 2012;82:e639-e645. 22. Rit S, van HM, Zijp L, et al. Quantification of the variability of diaphragm motion and implications for treatment margin construction. Int J Radiat Oncol Biol Phys 2012;82:e399-e407. 23. Guckenberger M, Meyer J, Wilbert J, et al. Intra-fractional uncertainties in cone-beam CT based image-guided radiotherapy (IGRT) of pulmonary tumors. Radiother Oncol 2007;83:57-64.