Subgroup Survival Analysis in Stage I-II NSCLC Patients With a Central Tumor Partly Treated With Risk-Adapted SBRT

International Journal of

Radiation Oncology biology

physics

www.redjournal.org

Clinical Investigation

Subgroup Survival Analysis in Stage I-II NSCLC Patients With a Central Tumor Partly Treated With Risk-Adapted SBRT Barbara Stam, PhD,* Margriet Kwint, MSc,* Matthias Guckenberger, MD, PhD,y,z Frederick Mantel, MD, PhD,y Andrew Hope, MD,x Meredith Giuliani, MD, PhD,x Maria Werner-Wasik, MD, PhD,k Inga Grills, MD, PhD,{ Jan-Jakob Sonke, PhD,* and Jose´ Belderbos, MD, PhD* *Department of Radiation Oncology, The Netherlands Cancer Institute, Amsterdam, The Netherlands; yDepartment of Radiation Oncology, University of Wuerzburg, Germany; zDepartment of Radiation Oncology, University Hospital Zurich, Zurich, Switzerland; xUniversity of Toronto and Princess Margaret Cancer Center, Toronto, Ontario, Canada; kDepartment of Radiation Oncology, Thomas Jefferson University Hospital, Philadelphia, Pennsylvania; and {Beaumont Hospital, Royal Oak, Michigan Received Nov 23, 2017. Accepted for publication Aug 24, 2018.

Summary We investigated survival in 765 patients with early-stage non-small cell lung cancer who were treated with stereotactic body radiation therapy. We distinguished between gross tumor volumes (GTVs) in the first or second centimeter surrounding the proximal bronchial tree and compared them with those in peripheral tumors. Patients with a GTV within

Purpose: Stereotactic body radiation therapy has been associated with increased toxicity when delivered to patients with early-stage non-small cell lung cancer with a tumor within 2 cm of the proximal bronchial tree (PBT). We investigated noncancer deaths for these patients as related to gross tumor volume (GTV) proximity to the PBT, compared with peripheral tumors. Methods and Materials: We included 765 patients with early-stage non-small cell lung cancer who were treated with stereotactic body radiation therapy to a median of 3  18 Gy. Central tumors were treated with a risk-adapted (less-intense) schedule (mostly 8 fractions) in 55% of the patients in the first-centimeter group and 27% of the patients in the second-centimeter group. An average anatomy with contouring of PBT and organs at risk (OARs) was deformed onto each patient to obtain the distance of the GTV to the PBT and doses to OARs. Log-rank, 1-way analysis of variance, and Cox regressions were performed to assess differences in the first-centimeter, second centimeter, and peripheral groups and associations with noncancer deaths.

Reprint requests to: Jose´ Belderbos, MD, PhD, Department of Radiation Oncology, The Netherlands Cancer Institute Plesmanlaan 121, 1066 CX Amsterdam, The Netherlands. Tel: þ31205122174; E-mail: [email protected]

Int J Radiation Oncol Biol Phys, Vol. 103, No. 1, pp. 132e141, 2019 0360-3016/$ - see front matter Ó 2018 Elsevier Inc. All rights reserved. https://doi.org/10.1016/j.ijrobp.2018.08.040

This research was partially supported by Elekta through a research grant, with all institutions being members of the Elekta Lung Research Group. This work and these data, however, are the intellectual property of the individual group members and their sponsoring institutions. Conflict of interest: none.

Volume 103  Number 1  2019

the first cm died 3 times more often from causes other than cancer compared with other patients. Noncancer deaths in patients with a GTV in the second centimeter, who partly received a riskadapted schedule, were comparable to those of patients with a peripheral GTV.

Risk-adapted SBRT for central tumors

133

Results: The median overall survival was 42.7 months, the median noncancer death occurred in 57.3 months, and the median follow-up was 34.8 months. Noncancer death in the first-centimeter group (31 patients) was significantly different from noncancer death in the other groups, with a hazard ratio of 3.175 (P < .001). Noncancer death in the second-centimeter group (71 patients) was not different from noncancer death in the peripheral group (P Z .53). Doses to OARs were higher in the first- and secondcentimeter groups than in the peripheral group for all OARs. High dose to the PBT was associated with noncancer death (D1%; hazard ratio, 1.006 Gy1; P Z .003). Conclusions: Patients with a GTV in the first centimeter surrounding the PBT died more often from causes other than cancer compared with other patients. Noncancer death in patients with a GTV in the second centimeter, who partly received a risk-adapted schedule, was comparable to that in patients with a peripheral tumor. Ó 2018 Elsevier Inc. All rights reserved.

Introduction Stereotactic body radiation therapy (SBRT) is the preferred curative treatment option for inoperable patients with peripheral, early-stage, non-small cell lung carcinoma (NSCLC).1,2 The high biological doses given with SBRT may, however, cause increased toxicity of the central structures (main airways, large blood vessels, heart, esophagus, and brachial plexus). In 2004, in a Japanese retrospective multicenter study in which a range of SBRT fractionation schedules was used, no difference in survival and toxicity was found.3 However, for patients with a central tumor, the Indiana phase 2 trial that served as a basis for the Radiation Therapy Oncology Group (RTOG) 0236 trial reported excessive toxicity for patients with stage I NSCLC with a centrally located tumor.4 Researchers from this trial advised that an SBRT regimen of 60 to 66 Gy in 3 fractions should not be used for patients with a tumor near the central airways.4 Since then, the general approach to avoid toxicities is to give a risk-adapted SBRT schedule using a more fractionated regimen (5-10 fractions) or to treat in the context of a clinical trial.5-10 Three phase 1 or 1 to 2 trials that use SBRT for centrally located lung tumors are currently accruing or awaiting final results.11-13 The definition of a central tumor varies in these trials, but are all based on the distance between the tumor and the proximal bronchial tree (PBT). The PBT includes the distal 2 cm of the trachea, the carina, the right and left mainstem bronchi, the right and left upper lobe bronchi, the intermedius bronchus, the right middle lobe bronchus, the lingular bronchus, and the right and left lower lobe bronchi. In the RTOG 0236 trial, “centrality” was defined as a tumor located within or touching the zone of the PBT, defined as a volume 2 cm in all directions around the PBT. The RTOG 0813 trial used the definition from RTOG 0236 and added tumors that were immediately adjacent to mediastinal or pericardial pleura. The European Organisation for Research and Treatment of Cancer (EORTC) LungTech trial defined “central” as in the RTOG 0236 or, if the tumor was immediately adjacent to the mediastinal or pericardial pleura, with a planning target volume (PTV)

expected to touch or include the pleura.12 The Nordic HILUS trial defined “central” as a tumor 1 cm from the PBT.11 Furthermore, there are important differences in dose prescription in the trials; RTOG 0813 is a trial with various dose levels, where dose was prescribed to the edge of the PTV.14 The EORTC LungTech trial prescribes 8  7.5 Gy to the PTV.12 The Nordic HILUS trial applies 8  7 Gy to the 65% to 70% isodose line.11 Preliminary results from this trial showed higher toxicity for patients with a central tumor close to the main bronchus compared with a lobar bronchus.11 Preliminary results from the RTOG 0813 trial showed comparable toxicities for central tumors compared with peripheral tumors.13,15 In this article, we investigate survival of patients with a tumor in the first centimeter surrounding the PBT or the second centimeter (the zone between 1 and 2 cm) and compare these results with survival in patients with peripheral tumors. We used deformable registration techniques to estimate PBT proximity for a large multiinstitutional data set of patients treated with a range of doses and fractionations. We aim to identify relationships between PBT proximity and survival outcomes unrelated to cancer recurrence.

Methods and Materials Patient characteristics Between 2005 and 2013, 1337 patients were treated for stage T1a to T2bN0 NSCLC with image-guided SBRT in 5 international institutes (The Netherlands Cancer Institute, Amsterdam, The Netherlands; William Beaumont Hospital, Royal Oak, Michigan; University of Wuerzburg, Wuerzburg, Germany; Princess Margaret Hospital, University of Toronto, Toronto, Ontario, Canada; and Thomas Jefferson University Hospital, Philadelphia, Pennsylvania). Eightyseven percent of patients were medically inoperable.16 All patients with a single tumor for whom SBRT treatment plans were available were included (765 patients; Table 1). We excluded patients who had a synchronous or

134

International Journal of Radiation Oncology  Biology  Physics

Stam et al.

Table 1

Patient and tumor characteristics

All No. Male (%) Median age (y) Median tumor diameter (cm) T-stage (%) T1 T2 Unknown Lobe (%) RUL RML RLL LUL Lingula LLL Histology (%) Adenocarcinoma Squamous Other Unknown WHO performance status (%) 0 1 2 3 Unknown Charlson comorbidity index (%) 0 1-2 3-4 5 Unknown FEV1 predicted (%) <80% 80% DCLO predicted mean (%) Fractionation schedule (%) 3  20 Gy 3  18 Gy 3  15 Gy 3  12.5 Gy 3  8 Gy 4  12 Gy 4  10.6 Gy 5  12 Gy 5  10 Gy 5  8 Gy 8  7.5 Gy

765 53 75 2.2

Table 1

(continued )

GTV edge GTV edge in first in second GTV edge centimeter centimeter peripheral 31 52 75 4.1

71 59 76 2.7

663 53 74 2.2

69 27 5

32 42 26

55 39 6

72 24 4

36 4 18 28 1 14

55 0 13 19 6 6

59 6 20 13 0 3

33 4 18 30 0 16

30 20 12 38

26 32 10 32

32 15 10 43

30 20 12 38

25 49 22 2 2

29 55 13 3 0

30 45 23 3 0

25 49 22 2 2

8 26 31 6 29

13 52 19 3 13

17 25 32 6 20

6 25 31 6 31

75 25 54

81 19 50

69 31 56

75 25 54

1 60 3 3 0 19 0 6 1 0 3

0 19 3 3 0 16 0 10 6 3 13

1 46 8 4 1 20 1 7 0 0 13

1 63 3 3 0 19 0 5 1 0 2 (continued)

All 8  6 Gy 10  5 Gy Other Median BED with a/b 10 Gy (Gy) Median mean lung dose (Gy) Median follow-up (mo) Median overall survival (mo) Median noncancer death (mo) Survival (%) Alive Noncancer death Cancer death Overall survival (%) 1 year 2 year 5 year Noncancer death (%) 1 year 2 year 5 year

1 1 1 151 4.29

GTV edge GTV edge in first in second GTV edge centimeter centimeter peripheral 19 10 0 105 6.46

0 1 0 132 5.30

0 0 1 151 4.13

34.8

75.9

36.1

34.5

42.7

13.7

54.2

44.3

57.3

15.6

66.6

58.7

58 31 11

19 58 23

58 30 13

60 30 11

86 68 37

57 23 6

86 68 44

88 70 38

90 75 47

65 33 10

91 76 58

91 77 48

Abbreviations: BED Z biologically effective dose; DLCO Z diffusion capacity of lung for carbon monoxide; FEV1 Z forced expiratory volume in 1 second; GTV Z gross tumor volume; LLL Z left lower lobe; LUL Z left upper lobe; RLL Z right lower lobe; RML Z right middle lobe; RUL Z right upper lobe; WHO Z World Health Organization.

metachronous tumor (293 patients) and those for whom we did not have a complete planning data set (279 patients). The different institutes used various treatment schedules. For tumors with the edge of the gross tumor volume (GTV) within the first centimeter surrounding the PBT, 8  6 Gy (19%), 3  18 Gy (19%), and 4  12 Gy (16%) were the dominant schedules (Table 1). All centers used a 4-dimensional planning computed tomography (CT) scan. Details for GTV, internal target volume (ITV), clinical target volume, and PTV were described previously.16 Inhomogeneous doses to the PTV to a maximum of 110% to 165% were allowed (depending on the center). Each patient was treated with 3-dimensional (3D) conformal radiation therapy, intensity modulated radiation therapy, or volumetric modulated arc therapy, using daily online cone beam CTebased volumetric image guidance. Each center had independent research ethics board approval for participation and analysis of collaboration data and outcomes.

Volume 103  Number 1  2019

Risk-adapted SBRT for central tumors

135

Follow-up consisted of physical examinations and CT scans at least every 6 months for the first 2 years and annually thereafter. Cause of death from patients with progressive disease, second cancers, or metastases at time of death was scored as death from cancer unless specified otherwise. All other causes of death, including unknown cause of death, were scored as noncancer death. Causes of death other than cancer or toxicity were available for 1 institute, in which 20% of causes of death were unknown.

Automatic segmentation We used an average anatomy to contour the organs at risk (OARs) in all patients. This average anatomy was built by using 109 patients from the 5 institutes and represents the average shape and intensity of these patients. A detailed description was previously reported.17 Within this average anatomy, a radiation oncologist and a physician assistant contoured the PBT, trachea, esophagus, and heart following RTOG 0813 and the RTOG lung atlas.18 The average anatomy was deformed onto each individual patient using ADMIRE (ADMIRE Research 2015, Elekta AB, Stockholm, Sweden), and all structures were deformed accordingly. Accuracy of deformation was assessed using absolute mean distance to agreement between manual and automatic PBT contours for 10 patients.17 The effect on the difference between manual and automatic contours in PBT to GTV distance was also assessed.

Survival stratified by PBT proximity For each patient, the shortest distance between PBT and the edge of the GTV was measured automatically using inhouse software (Match42). In 8 cases (1%), the GTV was not available; instead, the ITV was used. Patients were split into 3 groups according to the distance from the PBT to the GTV edge: first-centimeter (<1 cm), second-centimeter (1 cm, <2 cm), and peripheral (2 cm; Fig. 1). The difference among all groups in noncancer death and cancer death was tested using the log-rank test. The Kaplan-Meier noncancer death curve was plotted for the 3 groups (ie, death from cancer was censored). Second, to account for inherent survival differences because of tumor size, each group was split on the first-centimeter group’s median GTV, and Kaplan-Meier curves were plotted. Next, we assessed noncancer death for the central (first- and second-centimeter groups combined) versus the peripheral group. Within each group, we tested the difference in survival for patients who did or did not receive a riskadapted schedule using the log-rank test. We also tested the effect size for each of these groups using univariate Cox regressions. To compare our results with those of the phase 2 Nordic HILUS trial, we assessed noncancer death for patients with

Fig. 1. PBT (yellow) contoured in average anatomy, first cm surrounding PBT (green) and second cm surrounding PBT (green), lungs (blue). Abbreviation: PBT Z proximal bronchial tree. (A color version of this figure is available at https://doi.org/10.1016/j.ijrobp.2018.08.040.) the GTV edge in the first centimeter, split between tumors close to a main bronchus or a lobar bronchus. Finally, we performed a multivariate Cox regression in which we looked at the distance to the PBT in the 3 groups and added clinical factors that have a possible association with death: age, sex, tumor diameter, World Health Organization performance score, Charlson comorbidity index, lung function (forced expiratory volume in 1 second), T-stage, histology, pack-years of smoking, and mean lung dose.

Dose to OARs We used the PBT, trachea, esophagus, heart, and heart substructures from the automatic segmentations and the lungs-GTV contour from the original plans. The 3D dose distributions were converted voxel by voxel to biologically equivalent doses as given in fractions of 2 Gy using the linear-quadratic model with a/b Z 3 Gy to account for the dose per fraction.19,20 For each OAR except the heart and lungs, we calculated maximum dose and the V5Gy and V15Gy (the volume of structure receiving at least a 5- or 15Gy dose). For the heart, we also calculated the D1%, which was shown to be associated with noncancer death.17 For the esophagus, we also calculated the V50Gy, which was shown to be associated with acute esophageal toxicity,21 and the maximum dose. For the lungs, we calculated the mean lung dose and the V5Gy and V15Gy. To evaluate whether the doses on the OARs were different between the first- or second-centimeter groups and the peripheral group, we performed a 1-way analysis of variance. We also performed a series of univariate Cox

International Journal of Radiation Oncology  Biology  Physics

Stam et al.

regressions to test the association between PBT dose and noncancer death, for various values of DV (V range, 0%100%) and VD (D range, 0-55 Gy). The optimal model was selected based on the Akaike information criterion, and values for these dosimetric parameters were added to the dose table (Table 2). All tests for significance were conducted at the 5% significance level using SPSS (IBM SPSS, Version 22.0., Armonk, NY).

Table 2

Time trend To assess for a time bias in patients dying of causes other than cancer, we plotted a Kaplan-Meier curve for patients who died of such causes, split by treatment year: those treated before 2010 and those treated in or after 2010. A log-rank test was used to assess the difference. In the event of significant difference, we assessed noncancer deaths in the central and peripheral groups for the patients treated before 2010 versus those treated in or after 2010, using univariate Cox regression.

Median doses to organs at risk GTV edge in GTV edge first in second centimeter centimeter

Heart D1% (Gy) V5 (%) V15 (%) Lungs-GTV or lungs-ITV Mean lung dose (Gy) V5 (%) V15 (%) Esophagus Maximum dose (Gy) V5 (%) V15 (%) V50 (%) PBT Maximum dose (Gy) D1% (Gy) V5 (%) V15 (%) Main bronchus Maximum dose (Gy) V5 (%) V15 (%) Lobar bronchi Maximum dose (Gy) V5 (%) V15 (%) Trachea Maximum dose (Gy) V5 (%) V15 (%)

Results GTV edge peripheral

32.8* 21.5* 5.7*

19.1* 9.3 2.0

11.8 7.6 0.2

9.2*

7.6*

5.8

24.5* 13.1*

20.1* 10.8*

16.1 8.2

32.5*

20.7

15.2

31.4* 14.8* 0

25.6* 1.3 0

20.0 0.02 0

151.0*

71.1*

22.3

83.2* 48.2* 19.8*

44.3* 40.9* 14.1*

15.8 21.6 1.4

85.8*

43.8*

10.7

82.1* 40.3*

61.3* 21.8*

11.9 0

99.7*

68.0*

12.6

Validation of the deformable image registration resulted in a mean distance to agreement for the PBT contours of 1.3 mm (interquartile range, 1.1-1.4 mm). The median difference in GTV to PBT distance was 3.4 mm (interquartile range, 2.5-10 mm). Of the 765 patients, 31 had a tumor in the first centimeter surrounding the PBT (GTV edge <1 cm from the PBT) and 71 had a tumor in the second centimeter (GTV edge between 1 and 2 cm from PBT; Table 1). The median GTV diameter was 4.1 cm in the first-centimeter group, 2.7 cm in the second-centimeter group, and 2.2 cm in the peripheral group (displayed graphically in Fig. 2). Histology was available for 62% of patients; 30% were adenocarcinoma, 20% were squamous cell carcinoma, and 12% were other types of carcinoma. No significant differences in the groups were found other than tumor diameter. A riskadapted schedule was given to 55% of the patients in the first-centimeter group and to 27% of the patients in the second-centimeter group. The median biologically equivalent doses with a/b Z 10 Gy were 105 Gy in the firstcentimeter group, 132 Gy in the second-centimeter group, and 151 Gy in the peripheral group. The median follow-up was 34.8 months (calculated with the Kaplan-Meier method22), the median overall survival

6.00

44.4* 17.3*

26.0* 11.7*

10.6 0

17.5*

15.4*

5.5

12.7 0.6

14.9 0.02

0.3 0

Abbreviations: D1% Z dose to 1% of the volume of the structure; GTV Z gross tumor volume; ITV Z internal target volume; PBT Z proximal bronchial tree; V5 Z the volume of structure receiving at least a 5-Gy dose; V15 Z the volume of structure receiving at least a 15-Gy dose; V50 Z the volume of structure receiving at least a 50-Gy dose. * Doses in the first- or second-centimeter groups that were significantly different from those in the peripheral group.

Tumor diameter (cm)

136

4.00

2.00

.00

.00

8.00 10.00 2.00 4.00 6.00 Distance to proximal bronchial tree (cm)

12.00

Fig. 2. Tumor diameter as a function of distance to the proximal bronchial tree.

Volume 103  Number 1  2019

Risk-adapted SBRT for central tumors

was 42.7 months, and the median noncancer death occurred at 57.3 months. Death from cancer was significantly higher in the first-centimeter group than in the second-centimeter and peripheral groups (log-rank, P Z .019), with estimated effect size hazard ratios (HRs) of 3.132 (confidence interval [CI], 1.355-7.241) and 1.104 (CI, 0.55-2.217), respectively. The Kaplan-Meier curve for noncancer death for the 3 groups is plotted in Figure 3. Noncancer death in the firstcentimeter group was significantly different from noncancer death in the second-centimeter and peripheral groups (log-rank, P < .001). The estimated effect size HR for the first-centimeter group compared with the peripheral group was 3.175 (CI, 1.955-5.156). Noncancer death in the second-centimeter group was not different from noncancer death in the peripheral group (log-rank, P Z .532), with estimated effect size HR of 0.866 (CI, 0.551-1.361). Noncancer death for patients with a central tumor (firstand second-centimeter groups combined) was not significantly different from noncancer death for patients with a peripheral tumor (log-rank, P Z .071), with estimated effect size HR of 1.395 (CI, 0.998-1.950). Noncancer death for tumors larger and smaller than 4.1 cm (the median GTV diameter in the first-centimeter group) is shown in Figure 4. In the second-centimeter and peripheral groups, 16% and 8% had a tumor >4.1 cm, respectively. Noncancer death in the smaller-tumor firstcentimeter group was significantly different from 1.0

GTV < 1 cm from PBT GTV 1-2 cm from PBT GTV peripheral

1-cumulative Non-cancer death

0.8

0.6

0.4

0.2

0.0

1st cm: 2nd cm: Peri:

0

20

31 71 663

8 40 353

40 60 Follow-up (months) 4 1 20 8 153 59

80

100

120

0 5 11

Fig. 3. Kaplan-Meier curve for noncancer death for the first centimeter (black), the second centimeter (dark gray), and the peripheral (light gray) tumors. Logrank Z P < .001. Abbreviations: GTV Z gross tumor volume; PBT Z proximal bronchial tree; Peri Z peripheral tumor.

137

noncancer death in the peripheral group (log-rank, P < .001), with estimated effect size HR of 2.634 (CI, 1.392-4.986), and for the larger-tumor first-centimeter group (log-rank P < .001), with estimated effect size HR of 4.426 (CI, 1.958-10.002). In both the first- and second-centimeter groups, there was no significant difference in survival between the patients who did or did not receive a risk-adapted treatment schedule, with log-rank P values of .067 and .880, respectively. In the first-centimeter group, 5 patients had a tumor close to a main bronchus, and 26 patients had a tumor close to a lobar bronchus. Noncancer death was not significantly different between these groups (log-rank, P Z .051). Estimated effect size HR for tumors close to a lobar bronchus compared with the main bronchus was 0.167 (CI, 0.022-1.277). The multivariate Cox regression showed that the distance to the PBT in 3 groups was significantly associated with noncancer death (P < .001). The HR compared with peripheral tumors was 3.307 in the first-centimeter group and 0.749 in the second-centimeter group. It was also associated with age (P Z .002; HR, 1.025), WHO performance status (P <.001; HR compared with WHO status 0 and WHO status 1, 1.617; WHO status 2, 3.410), lung function (P Z .007; HR, 0.991), and T-stage (P, .019; HR, 1.076). Doses to the heart, esophagus, and PBT were 3, 2, and 7 times higher, respectively, in the first-centimeter group than in the peripheral group (Table 2). In both the first- and the second-centimeter groups, most doses were significantly higher than in the peripheral group. The series of univariate Cox regressions showed significant associations between PBT dose and noncancer death for the high V values (V40Gy-V55Gy) and most D values (D0%-D80%). The strongest association was found for the D1%, with an HR of 1.006 Gy1 (CI, 1.002-1.011) and P value of .003. For patients treated before 2010, the median follow-up was 66.9 months, the median overall survival was 35.6 months, and the median noncancer death occurred at 50.9 months. For patients treated in or after 2010, the median follow-up was 24.3 months, the median overall survival was 58.8 months, and the median noncancer death occurred at 58.8 months (Fig. 5). A log-rank test showed a significant difference in noncancer deaths between these groups (P value Z .002). Noncancer death was not significantly different for the central groups for patients treated before or during/after 2010. For patients with a peripheral tumor treated before 2010, the estimated effect size HR was 1.1613 (1.189-2.189; P Z .002) compared with patients treated in or after 2010.

Discussion We showed that death from causes other than cancer was 3.3 times higher for patients with a tumor in the first

138

International Journal of Radiation Oncology  Biology  Physics

Stam et al.

A 1.0

B

GTV < 1 cm from PBT

1.0

GTV < 1 cm from PBT

GTV 1-2 cm from PBT GTV peripheral

GTV 1-2 cm from PBT GTV peripheral 0.8 1-cumulative non-cancer death

1-cumulative non-cancer death

0.8

0.6

0.4

0.6

0.4

0.2

0.2

0.0

0.0 0

20

40

60

80

100

0

20

Follow up (months) 1st cm: 2nd cm: Peri:

15 10 50

2 6 30

1 5 12

0 2 5

40

60

80

100

120

Follow up (months) 0 2 2

1st cm: 16 2nd cm: 61 613 Peri:

6 35 322

3 15 141

1 6 54

0 3 9

Fig. 4. Kaplan-Meier curve for noncancer death, for the first centimeter (black), the second centimeter (dark gray), and the peripheral (light gray) tumors, for tumors larger (A) or smaller (B) than 4.1 cm in diameter. Log-rank Z P < .001 (A) and P < .001 (B). Abbreviations: GTV Z gross tumor volume; PBT Z proximal bronchial tree; Peri Z peripheral tumor.

centimeter surrounding the PBT than for patients with a peripheral tumor. For patients with a tumor in the second centimeter from the PBT, noncancer death was not significantly different from that in patients with a peripheral tumor. This effect was not due to the larger tumors in the first-centimeter group; Figure 4 shows that larger and smaller first-centimeter tumors both have worse outcomes compared with peripheral tumors. When we grouped the first- and second-centimeter tumors into a “central” group (following the RTOG 0236 definition of centrality), noncancer death was not significantly different from that in the peripheral group, which is comparable to results found in the literature.23 The tumors in the central group were larger than tumors in the peripheral group, likely because (1) most of these tumors are found incidentally on a 2-dimensional lung x-ray, on which peripheral tumors are more visible and therefore found earlier, and (2) for 2 tumors with the same center of mass, the larger tumor is inherently closer to the PBT. Because tumors were larger in the firstcentimeter group, the first-centimeter group contained more T2 tumors than the peripheral group. The higher death rates from cancer in the first-centimeter group can be attributed to this higher T2 stage and lower biologically effective dose.24 The physical condition of the patients in the firstcentimeter group was, on average, poorer than that of the patients with peripheral tumors, as witnessed by the higher percentage of patients with a forced expiratory volume in

1 second < 80% and a lower diffusion capacity of lung for carbon monoxide, World Health Organization performance status, and Charlson comorbidity index. The multivariate analysis, however, has shown that poorer physical shape alone does not explain the higher noncancer deaths in these patients and that distance to the PBT also plays a role. We did not include dose to the PBT in the multivariate analysis because it is strongly correlated with distance to the PBT. Another difference between the central and peripheral groups is the underlying physiology; in the first-centimeter group, squamous cell carcinoma was dominant, whereas adenocarcinoma was dominant in the second-centimeter and the peripheral groups. Because overall survival has been shown to be better for patients with adenocarcinoma,25 this factor could contribute to the survival difference between the groups. However, histology was not significantly associated with survival in the multivariate analysis. T-stage did show a significant association with noncancer death in the multivariate analysis. This association may be due to the estimated 20% of patients with an unknown cause of death, whose results were scored as noncancer death. Within the consortium, cause-specific survival was scored, not cause of death; it was therefore not possible to exclude patients with an unknown cause of death. Because death from cancer can be viewed as a competing risk for “developing” noncancer death, the inclusion of cancer deaths in the noncancer group would cause an underestimation of the true effect and the resulting T-stage in the multivariate analysis.

Volume 103  Number 1  2019

139

In 2014, the safety of using SBRT for central tumors was proclaimed when Chang et al compared survival for patients with a central tumor (using the definition of centrality from RTOG 0236) with that of patients with a peripheral tumor.10 These patients were treated with the same nonadapted schedule of 50 Gy in 4 fractions, and no difference in survival was found. Our results are comparable because (1) most patients did not receive a risk-adapted schedule and (2) when we combined first-and second-centimeter groups into “central” tumors, the rate of noncancer death between central and peripheral tumors was not different, despite significant differences in doses to the OARs. However, it is clear that patients with a tumor in the firstcentimeter surrounding the PBT do significantly worse. In the absence of prospective trial results demonstrating safety, these patients should preferably be treated within the context of a trial. Alternatively, recent American Society for Radiation Oncology guidelines suggest a risk-adapted schedule for these patients.34 As Chang et al10 have suggested, normal tissue constraints should lead in the fractionation scheme of SBRT, not the distance to PBT. The “central” definition of 2 cm surrounding the PBT could act as a surrogate for sparing large blood vessels, the heart, or the esophagus, structures for which it has been shown that dose is correlated to toxicity and survival in patients with early-stage NSCLC.17,21 The cohort was treated between 2005 and 2013, an era in which delivery techniques evolved from 3D conformal radiation therapy to intensity modulated radiation therapy and volumetric modulated arc therapy and where the risk of radiation pneumonitis could be reduced because of these newer techniques.35 Figure 5 shows that the patients treated in recent years have significantly greater survival, which is

1.0

treated before 2010 treated in or after 2010

0.8 1-cumulative non cancer death

In several retrospective studies, results on toxicity and survival vary. Comparing central to peripheral tumors, toxicity and survival were not different in 4 studies with 251, 100, 80, and 34 central tumors,10,23,26,27 but they were different in 2 studies with 108 and 90 central tumors.28,29 All of these retrospective studies used the RTOG definition for a central tumor of GTV <2 cm from the PBT or mediastinal envelope, following either RTOG 0236 or 0813.13,30 We hypothesize that the opposing results may arise from differences between tumors very close to the PBT and tumors farther away from the PBT (but still within 2 cm). We need to distinguish between central tumors <1 cm from the PBT and central tumors between 1 and 2 cm from the PBT. Two papers that report on tumors <1 cm from the PBT show higher toxicities.11,31 The higher grade 4 and 5 toxicity found in the Nordic HILUS trial for patients with a tumor close to the main bronchus compared with the lobar bronchi11 was not found in our data set (P Z .051), which was likely too small to find such differences. Also, our study differs in prescription dose; the other study gave 8  7.5 Gy. The literature shows 2 different definitions of the term “ultracentral” tumor: the GTV directly abutting the central airway27 or the PTV overlapping the trachea or main bronchi.31,32 The tumors in our first-centimeter group are generally considered ultracentral, but we refrain from using this term to avoid confusion: Our first-centimeter group also included tumors not abutting the central airway and tumors close to the lobar bronchi. Therefore, we find that a direct comparison to papers reporting on patients with an ultracentral tumor would be unfair, although both our results and the literature show worse toxicity or survival for first-centimeter/ultracentral tumors.11,31 Not all patients with a central tumor received a riskadapted schedule, for a variety of reasons. First, the SBRT dose prescription is estimated based on the diagnostic CT (a breath-hold CT), whereas the plan is made on a 4-dimensional CT. The tumor might be larger or located more closely to the central structures than anticipated. For some patients, although the center of mass of the tumor is peripheral, a small limb of the tumor protruded toward the central zone. If this was not anticipated, the central bronchial tree might not have been delineated. A similar situation occurs when a lobar proximal bronchus extends more peripherally into the lung parenchyma and thus is closer to a peripheral tumor. This situation could have caused higher than anticipated doses on central structures. Second, all patients with a tumor in the first centimeter surrounding the PBT were treated before June 2013, of which 61% were treated before 2010 (and thus before a risk-adapted schedule was recognized as the preferred treatment after systematic review).33 We do not have information on how often a schedule was adapted during the planning because OAR dose constraints could not be met, nor do we have data on how often an underdosage of the PTV was accepted to meet OAR constraints.

Risk-adapted SBRT for central tumors

0.6

0.4

0.2

0.0 .00

20.00

40.00

60.00

80.00

100.00

120.00

Follow-up (months)

Fig. 5. Kaplan-Meier curve for noncancer death, for patients treated before 2010 (gray), or in/after 2010 (black).

140

Stam et al.

attributable to the patients with a peripheral tumor; patients with a central tumor did not show a difference in noncancer death. The effect of the evolved treatment techniques did not show a significantly better survival for the patients with a central tumor, which could be because the cohort is underpowered (19 patients vs 12 patients with a tumor in the first centimeter, treated before or during/after 2010, respectively). One of the limitations of this study is that we did not analyze the toxicity of central structures for this patient group separately because these data were not scored separately within the consortium. In the absence of prospective trials to report toxicity, we rely on reports of toxicity from retrospective studies. A wide range of toxicities in patients treated with SBRT has been reported: fatal hemoptysis, respiratory failure, atrioventricular block, bleeding, chestwall pain, dyspnea, empyema, fatigue, fever, fistula, lung infection, pain, pneumonitis, pneumothorax, and ventricular arrhythmia.11,31,36 The literature shows that toxicity is higher for ultracentral tumors,11,31 and it is ambiguous about central tumors.23,27-29 No data on toxicity for tumors specifically in the second centimeter from the PBT are available, to our knowledge. Other types of toxicity, not specifically related to the central structures, were reported previously for this cohort16; they include grade 2 or higher pneumonitis (7%), rib fracture (3%), myositis (1%), and dermatitis (2%). The local control of SBRT for patients with early-stage lung cancer is excellent (88% at 3 years37), so treatment-related toxicity is an important factor to consider. Guidelines for OARs, expected from ongoing trials, will facilitate reduction of toxicity for these patients. A second limitation is uncertainty in the cause of death. These patients were elderly (median age, 75 years) and had lung cancer and multiple comorbidities that commonly precluded surgery. Differentiation between treatmentrelated death and death resulting from comorbidities is therefore challenging, even within the framework of prospective trials, as argued by Guckenberger5 based on the results by Timmerman et al,4 but this should be further explored. Many patients died at home, and death certificates are known to be less accurate for patients who die outside of a hospital.38 Cause of death was available for 1 institute, but for 20% of the cases, the cause of death was unknown. Cause-specific survival was scored accurately for all institutes, with only 0.5% considered “unknown” cases. Thus, adding unknown causes of death in the “noncancer death” category is considered fair, and it does not cause an overestimation of the number of patients dying of causes other than cancer. A third limitation is the registration inaccuracy of the deformation of the PBT onto the individual patient of 1.3 mm, onto the individual patient, which could have caused a patient to be classified into the wrong group. However, a visual inspection of data from all patients in the firstcentimeter group did not show inaccurate registrations, and therefore we expect the effect of misclassification to have been minimal.

International Journal of Radiation Oncology  Biology  Physics

Fourth, the delineation of the GTV and expansion to ITV, clinical target volume, and finally PTV was institution dependent. We circumvented this issue by focusing on the distance between the GTV and PBT (as is the definition of centrality) and by looking at planned dose to the OARs.

Conclusions In this data set of 765 patients with early-stage NSCLC who were treated with SBRT, we showed that patients with a tumor in the first centimeter surrounding the PBT more often die of other causes than cancer. For patients with a tumor in the second centimeter surrounding the PBT, survival was comparable to that in patients with a peripheral tumor.

References 1. Timmerman RD, Paulus R, Galvin JM, et al. Stereotactic body radiation therapy for inoperable early stage lung cancer. JAMA 2010;303: 1070-1076. 2. Lagerwaard FJ, Verstegen NE, Haasbeek CJA, et al. Outcomes of stereotactic ablative radiotherapy in patients with potentially operable stage I non-small-cell lung cancer. Int J Radiat Oncol Biol Phys 2011; 83:348-353. 3. Onishi H, Araki T, Shirato H, et al. Stereotactic hypofractionated highdose irradiation for stage I nonsmall cell lung carcinoma: Clinical outcomes in 245 subjects in a Japanese multiinstitutional study. Cancer 2004;101:1623-1631. 4. Timmerman RD, McGarry RC, Yiannoutsos CT, et al. Excessive toxicity when treating central tumors in a phase II study of stereotactic body radiation therapy for medically inoperable early-stage lung cancer. J Clin Oncol 2006;24:4833-4839. 5. Guckenberger M. Dose and fractionation in stereotactic body radiation therapy for stage I non-small cell lung cancer: Lessons learned and where do we go next? Int J Radiat Oncol Biol Phys 2015; 93:765-768. 6. Nestle U, Belderbos J. Cons: Should a medically inoperable patient with a T2N0M0 non-small cell lung cancer central in the lung hilus be treated using stereotactic body radiotherapy? Transl Lung Cancer Res 2015;4:623-626. 7. Woodford K, Senthi S. Pros: Should a medically inoperable patient with a T2N0M0 non-small cell lung cancer central in the lung hilus be treated using stereotactic body radiotherapy? Transl Lung Cancer Res 2015;4:619-622. 8. Nestle U, Belderbos J. Rebuttal from Dr Nestle and Dr Belderbos. Transl Lung Cancer Res 2015;4:629. 9. Woodford K, Senthi S. Rebuttal from Ms Woodford and Dr Senthi. Transl Lung Cancer Res 2015;4:627-628. 10. Chang JY, Li QQ, Xu QY, et al. Stereotactic ablative radiation therapy for centrally located early stage or isolated parenchymal recurrences of non-small cell lung cancer: How to fly in a “no fly zone” Int J Radiat Oncol Biol Phys 2014;88:1120-1128. 11. Lindberg K, Bergstro¨m P, Brustugun OT, et al. OA24.05 The Nordic HILUS-Trial d First report of a phase II trial of SBRT of centrally located lung tumors. J Thorac Oncol 2017;12:S340. 12. Adebahr S, Collette S, Shash E, et al. LungTech, an EORTC phase II trial of stereotactic body radiotherapy for centrally located lung tumours: A clinical perspective. Br J Radiol 2015;88:20150036. 13. Bezjak A, Paulus R, Gaspar LE, et al. Primary study endpoint analysis for NRG Oncology/RTOG 0813 trial of stereotactic body radiation therapy (SBRT) for centrally located non-small cell lung cancer (NSCLC). Int J Radiat Oncol Biol Phys 2016;94:5-6.

Volume 103  Number 1  2019 14. Bezjak A. RTOG 0813: Seamless phase I/II study of stereotactic lung radiotherapy (SBRT) for early stage, centrally located, non-small cell lung cancer (NSCLC) in medically inoperable patients. RTOG 0813 2011;1-80. Philadelphia: RTOG. Available at: https://www.rtog.org/ ClinicalTrials/ProtocolTable/StudyDetails.aspx?studyZ. 15. Bezjak A, Paulus R, Gaspar LE, et al. OC-0136: Primary study endpoint analysis of NRG Oncology/RTOG 0813 trial of SBRT for centrally located NSCLC. Radiother Oncol 2016;119:S61-S62. 16. Grills IS, Hope AJ, Guckenberger M, et al. A collaborative analysis of stereotactic lung radiotherapy outcomes for early-stage non-small-cell lung cancer using daily online cone-beam computed tomography. J Thorac Oncol 2012;7:1382-1393. 17. Stam B, Peulen H, Guckenberger M, et al. Dose to heart substructures is associated with non-cancer death after SBRT in stage I-II NSCLC patients. Radiother Oncol 2017;123:370-375. 18. Kong F-M, (Spring), Ritter T, Quint DJ, et al. Consideration of dose limits for organs at risk of thoracic radiotherapy: Atlas for lung, proximal bronchial tree, esophagus, spinal cord, ribs, and brachial plexus. Int J Radiat Oncol 2011;81:1442-1457. 19. Lebesque JV, Keus RB. The simultaneous boost technique: The concept of relative normalized total dose. Radiother Oncol 1991;22: 45-55. 20. Joiner MC, Bentzen SM. Fractionation: The linear quadratic approach. In: Basic Clinical Radiobiology; Great Britain: Hodder Arnold; 2009; p. 102-119. 21. Kwint M, Uyterlinde W, Nijkamp J, et al. Acute esophagus toxicity in lung cancer patients after intensity modulated radiation therapy and concurrent chemotherapy. Int J Radiat Oncol Biol Phys 2012;84:223-228. 22. Schemper M, Smith TL. A note on quantifying follow-up in studies of failure time. Control Clin Trials 1996;17:343-346. 23. Tekatli H, Senan S, Dahele M, et al. Stereotactic ablative radiotherapy (SABR) for central lung tumors: Plan quality and long-term clinical outcomes. Radiother Oncol 2015;117:64-70. 24. Zhang J, Yang F, Li B, et al. Which is the optimal biologically effective dose of stereotactic body radiotherapy for stage I non-smallcell lung cancer? A meta-analysis. Int J Radiat Oncol Biol Phys 2011; 81:305-316. 25. Asamura H, Goya T, Koshiishi Y, et al. A Japanese lung cancer registry study: Prognosis of 13,010 resected lung cancers. J Thorac Oncol 2008;3:46-52. 26. Park HS, Harder EM, Mancini BR, et al. Central versus peripheral tumor location: Influence on survival, local control, and toxicity

Risk-adapted SBRT for central tumors

27.

28.

29.

30.

31.

32.

33.

34.

35. 36.

37.

38.

141

following stereotactic body radiotherapy for primary nonesmall-cell lung cancer. J Thorac Oncol 2015;10:832-837. Chaudhuri AA, Tang C, Binkley MS, et al. Stereotactic ablative radiotherapy (SABR) for treatment of central and ultra-central lung tumors. Lung Cancer 2015;89:50-56. Haseltine JM, Rimner A, Gelblum DY, et al. Fatal complications after stereotactic body radiation therapy for central lung tumors abutting the proximal bronchial tree. Pract Radiat Oncol 2016;6:e27-e33. Schanne DH, Nestle U, Allga¨uer M, et al. Stereotaktische Strahlentherapie bei zentralem NSCLC im Stadium I: Eine multizentrische Analyse. Strahlenther Onkol 2014;191:125-132. Timmerman RD, Michalski J, Fowler J, et al. Radiation Therapy Oncology Group RTOG 0236: A phase II trial of stereotactic body radiation therapy (SBRT) in the treatment of patients with medically inoperable stage I/II non-small cell lung cancer. RTOG 2004;6:1-44. Philadelphia: RTOG. Available at: https://www.rtog.org/Clinical Trials/ProtocolTable/St. Tekatli H, Haasbeek N, Dahele M, et al. Outcomes of hypofractionated high-dose radiotherapy in poor-risk patients with “ultracentral” non-small cell lung cancer. J Thorac Oncol 2016;11:1081-1089. Baker S, Dahele M, Lagerwaard FJ, et al. A critifcal review of recent developments in radiotherapy for non-small cell lung cancer. Radiat Oncol 2016;11:115. Senthi S, Haasbeek CJA, Slotman BJ, et al. Outcomes of stereotactic ablative radiotherapy for central lung tumours: A systematic review. Radiother Oncol 2013;106:276-282. Videtic GMM, Donington J, Giuliani M, et al. Stereotactic body radiation therapy for early-stage non-small cell lung cancer: Executive summary of an ASTRO evidence-based guideline. Pract Radiat Oncol 2017;7:295-301. Hu X, He W, Wen S, et al. Is IMRT superior or inferior to 3DCRT in radiotherapy for NSCLC? A meta-analysis. PLoS One 2016;11:1-15. Oskan F, Becker G, Bleif M. Specific toxicity after stereotactic body radiation therapy to the central chest. Strahlenther Onkol 2017;193: 173-184. Fakiris AJ, McGarry RC, Yiannoutsos CT, et al. Stereotactic body radiation therapy for early-stage non-small-cell lung carcinoma: Fouryear results of a prospective phase II study. Int J Radiat Oncol Biol Phys 2009;75:677-682. Johansson LA, Westerling R. Comparing Swedish hospital discharge records with death certificates: Implications for mortality statistics. Int J Epidemiol 2000;29:495-502.