Abstract
Introduction
To delineate an optimal clinical target volume (CTV) for gastroesophageal junction (GEJ) cancers by comparing locoregional vs. extended locoregional radiation volumes.
Materials
This retrospective analysis examined 222 patients (111 matched pairs treated with locoregional vs. extended locoregional radiation) with non-metastatic GEJ carcinomas treated with concurrent chemoradiation +/- surgery. The CTV for locoregional volumes was defined as gross tumor volume (GTV) + elective coverage of first-echelon nodes and sometimes the celiac axis. The CTV for extended locoregional volumes was defined as GTV + elective coverage of celiac and splenic (+/- porta) nodes. Variables used for matching included gender, stage, performance status, histology, receipt of induction chemotherapy, type of concurrent chemotherapy, radiation modality, receipt of surgery, type of surgery, and Siewert classification. Regression models stratified by matched pairs were fit to estimate effect of radiation volume on clinical endpoints.
Results
Adjusting p-values for multiple testing, patients treated with extended locoregional vs. locoregional radiation had increased odds of grade 2+ acute chemoradiation-associated GI toxicity (OR 2.92, adj. p=0.0447). However, differing radiation volumes were not significantly associated with postoperative complication rates, pathologic T-stage, frequency of positive perigastric/regional nodes on surgical specimen, distant-metastases progression free survival, locoregional progression free survival, or overall survival (adj. p>0.05). Of the patients who did (N=124) and did not (N=72) receive elective celiac radiation, 2 (1.6%) and 2 (2.8%) patients, respectively, relapsed in the celiac nodes. No patients failed in the splenic or porta nodes.
Conclusions
Most GEJ cancers can be safely treated without elective inclusion of splenic/porta nodes.
Keywords: gastroesophageal cancer, radiation volume
Introduction
While radiation in the management of non-metastatic esophageal cancer has led to significant improvements in survival1-4, it remains unclear how to define the optimal CTV, especially for GEJ carcinomas. Part of the uncertainty in target definition arises from the controversy regarding whether GEJ carcinomas should be classified as gastric or esophageal carcinomas5-9. Although the most recent AJCC classification changed the categorization of all tumors located within 5cm below the GEJ and infiltrating the junction as esophageal cancers, several studies suggest that GEJ carcinomas should be more appropriately classified as gastric cancers7-9.
The challenge with this uncertainty in classification is that the recommended target volumes for gastric cancers differ from that recommended for esophageal cancers. Based on National Comprehensive Cancer Network (NCCN) guidelines, the CTV for GEJ esophageal tumors should include elective coverage of the first-echelon nodes including para-esophageal, lesser curvature, and celiac axis. For more gastric-centric tumors at the GEJ, however, NCCN also suggests additional CTV coverage of regional splenic/porta nodes if there's extension into or involvement of the body/middle-third of the stomach.
Questions thus remain regarding the optimal CTV for GEJ carcinomas – should these tumors be contoured liked esophageal tumors with locoregional radiation volumes (coverage of first-echelon nodes +/- celiac axis), or like gastric tumors with more extensive locoregional radiation volumes (additional splenic and porta nodal coverage)? Advocates of extended locoregional nodal radiation are concerned about the risk of nodal relapse, while opponents of extended locoregional nodal radiation are concerned about the risk of toxicity with bigger radiation fields. Large-scale randomization of differing contouring volumes in a prospective clinical trial would be difficult given the inability to adequately control for the numerous prognostic covariates and diverse treatment options. To assess the impact of additional elective splenic/porta nodal radiation on patients receiving chemoradiation for non-metastatic GEJ carcinomas, we retrospectively compared locoregional vs. extended locoregional radiation (RT) volumes in terms of acute toxicities, postoperative complications, patterns of failure, and survival.
Materials and Methods
Patients
This retrospective analysis was approved by the Institutional Review Board. We initially identified 824 patients with non-metastatic GEJ carcinomas treated at a single institution with concurrent chemoradiation +/- surgery, between 1998 and 2013. For consistency, staging was determined according to the 6th (2002) edition of the AJCC staging manual, using a combination of CT, PET/CT, and EGD/EUS with FNA biopsy of suspicious lymph nodes. Most patients were treated with locoregional radiation (N=707) and fewer patients were treated with extended locoregional radiation (N=117).
Esophageal cancer was defined per AJCC 7th edition staging10 as 15cm from the incisors to the GEJ and the proximal 5cm of the stomach. A tumor was defined as esophageal (rather than gastric) carcinoma if its epicenter was in the lower thoracic esophagus or GEJ, or within the proximal 5cm of the stomach with the tumor mass extending into the GEJ or distal esophagus. If the epicenter of the tumor was >5cm distal to the GEJ, or within 5cm of the GEJ but with no extension into the GEJ/esophagus, then the tumor was classified as gastric carcinoma and excluded from analysis. Only patients with GEJ esophageal carcinomas (including Siewert types I-III) were analyzed for this study.
Treatment
The CTV for locoregional radiation volumes was defined as GTV (tumor + involved nodes) + elective coverage of first echelon nodes and sometimes the celiac axis + 3-4cm superior/inferior margin + 1.0-1.5cm radial margin. In contrast, the CTV for extended locoregional radiation volumes was defined as GTV + 3cm mucosal expansion + elective coverage of celiac and splenic (+/- porta) nodes + additional 1.5cm superior/inferior and 1.0cm radial margin. The planning target volume (PTV) used for statistical analyses for both RT groups was defined as CTV + 5mm. Locoregional radiation was delivered to a median dose of 50.4 Gy in 28 fractions, while extended locoregional radiation was delivered to a median dose of 45 Gy in 25 fractions. The radiation treatment plan of each study patient was reviewed by a radiation oncologist (JW) to ensure consistency with the volumes described above. Illustrative examples of differences in locoregional vs. extended locoregional radiation volumes are shown in Figure 1. Radiation was delivered with 3-dimensional conformal radiation (3D-CRT), intensity-modulated radiation (IMRT), or proton beam therapy (PBT) and was based on physician preference.
Figure 1.
Representative A) extended locoregional radiation volumes with elective splenic +/- porta nodal coverage (blue line demarcates the PTV), and B) locoregional radiation volumes without elective splenic or porta nodal coverage (white line demarcates the PTV).
Chemotherapy agents consisted of fluoropyrimidine (IV or oral), given alone or in combination with either a platinum compound (classified as FP) or a taxane (classified as FT). Types of surgical procedures included Ivor-Lewis esophagectomy (with proximal gastrectomy and mediastinal + abdominal lymph node dissection), transthoracic esophagectomy, trans-hiatal esophagectomy, three-field esophagectomy, and minimally invasive esophagectomy, as well as total and subtotal gastrectomy (with esophageal margin and often modified D2 lymph node dissection for Siewert type III tumors). Of the 411 patients in the initial dataset (n=824) who received trimodality therapy, most patients (supplemental Table 1) treated with locoregional radiation had Ivor-Lewis (81%), followed by three-field (8%), trans-hiatal (7%) and minimally invasive esophagectomies (3%). In contrast, most patients treated with extended locoregional radiation had Ivor-Lewis esophagectomies (47%) followed by total gastrectomies (41%). Differences in surgery type were likely related to referral patterns, surgeon preference, radiation volume, and location of tumor/extent of gastric involvement, among others.
Definitions of clinical endpoints
All acute toxicities were graded according to Version 4.0 of the NCI Common Terminology Criteria for Adverse Events and assessed on a weekly basis during chemoradiation treatment. All postoperative complications were recorded up to 30 days postoperatively. Postoperative GI complications were defined as anastomotic leak, ileus, fistula, obstruction, or need for J-tube placement. Postoperative pulmonary complications included pneumonia, ARDS, pleural effusions, and/or respiratory insufficiency. Cardiac complications included atrial fibrillation, any non-specified arrhythmia, myocardial infarction, or development of congestive heart failure. Wound complications included wound erythema/infection (requiring antibiotic treatment or re-opening/packing of wound) or dehiscence.
Dates of death were determined by reviewing clinical follow-up information in the patients' medical records and Social Security Death Index. Overall survival (OS) was calculated from date of diagnosis to date of last known vital sign. Patients alive at last follow-up date were administratively censored. Hospital records from radiographic studies, follow-up clinical examinations, surgical explorations, and endoscopy at every 3 month to 1 year intervals were used to assess disease progression, locoregional failure, and distant failure. Distant metastases progression-free survival (PFS) was computed from the date of diagnosis to date of distant metastasis or death or last evaluation. Locoregional PFS was computed from the date of diagnosis to date of locoregional failure or death or last evaluation. Median follow-up time was comparable for both groups (23 months).
Statistical methods and matching
Data were collected retrospectively. Associations between categorical variables and between continuous measures and categorical variables were assessed using either Fisher's exact or generalized Fisher's exact tests and Wilcoxon rank sum test.
To correct for bias, starting with the initial dataset of 824 patients, a subset of 222 patients were chosen consisting of 111 matched pairs, one in each pair treated with locoregional and the other with extended locoregional radiation. The pairs were obtained using a nearest neighbor matching method11, 12. Variables used for matching included gender, tumor stage, performance status, histology, induction chemotherapy, type of concurrent chemotherapy, radiation modality, receipt of surgery, and type of surgery. Median year of therapy for both matched groups was 2006. Nearest neighbor matching also was used to produce a second dataset of 52 (locoregional vs. extended locoregional) pairs, matched by type II Siewert classification in addition to all previous covariates. matching by type I and III Siewert classification could not be performed due to the rarity of these tumors among patients treated with extended locoregional and locoregional radiation, respectively. Regression models stratified by matched pairs were fit to estimate the effect of locoregional vs. extended locoregional radiation on acute GI toxicity and postoperative complications (stratified logistic regression), and each type of PFS measure and OS (Cox proportional hazards regression models stratified by baseline hazard). For matched pairs of binary variables, association with locoregional vs. exended locoregional radiation was assessed using the McNemar mid-p test. Because many tests were performed, to control the overall false positive rate, all nominal p-values were adjusted for multiple testing via the Hochberg method.
Results
Patient characteristics
Table 1 summarizes baseline characteristics for all initial 824 GEJ patients and by locoregional vs. extended locoregional radiation volumes. Table 2 gives the corresponding summary for the 111 (locoregional vs. extended locoregional) matched pairs, and shows that none of the baseline covariates were statistically significantly different in the matched pair dataset. Among the 111 patients treated with extended locoregional radiation, 7 (6.3%), 51 (45.9%), and 53 (47.7%) patients had type 1, 2, and 3 Siewert GEJ carcinomas, respectively. Among the 111 patients treated with locoregional radiation, 63 (56.8%), 47 (42.3%), and 1 (0.9%) patient had types 1, 2, and 3 Siewert GEJ carcinomas, respectively.
Table 1. Baseline clinical and treatment characteristics for all patients and by extended locoregional vs. locoregional RT.
| Type of Radiation Volume | ||||
|---|---|---|---|---|
|
|
||||
| Measure | All Patients (N=824) | Extended Locoregional (N=117) | Locoregional (N=707) | Unadjusted P-Valuea |
| Age at diagnosis (years) | ||||
| Mean | 62.6 | 61.4 | 62.9 | 0.42b |
| Minimum, Maximum | 21.0, 91.0 | 21.0, 85.0 | 22.0, 91.0 | |
|
| ||||
| Gender, n (%) | ||||
| Male | 725 (88) | 93 (79) | 632 (89) | 0.0051c |
|
| ||||
| Tumor stage, n (%) | ||||
| 1 | 16 (2) | 1 (1) | 15 (2) | 0.0045 |
| 2 | 275 (34) | 26 (23) | 249 (36) | |
| 3 | 449 (56) | 82 (71) | 367 (53) | |
| 4a | 64 (8) | 6 (5) | 58 (8) | |
| Missing | 20 | 2 | 18 | |
|
| ||||
| Tumor length (cm) | ||||
| N | 714 | 90 | 624 | 0.56b |
| Mean | 5.4 | 5.6 | 5.4 | |
| Minimum, Maximum | 1.0, 20.0 | 1.0, 13.0 | 1.0, 20.0 | |
|
| ||||
| Karnofsky performance status | ||||
| Mean | 84.5 | 85.6 | 84.3 | 0.08b |
| Minimum, Maximum | 50.0, 100.0 | 60.0, 100.0 | 50.0, 100.0 | |
|
| ||||
| Number of comorbidities | ||||
| Mean | 1.1 | 1.1 | 1.1 | 0.94b |
| Minimum, Maximum | 0, 5.0 | 0, 4.0 | 0, 5.0 | |
|
| ||||
| Tumor differentiation, n (%) | ||||
| Well differentiated | 6 (1) | 1 (1) | 5 (1) | 0.58 |
| Moderate | 324 (40) | 42 (37) | 282 (40) | |
| Poor | 485 (60) | 72 (63) | 413 (59) | |
| Missing | 9 | 2 | 7 | |
|
| ||||
| Tumor histology, n (%) | ||||
| Adeno | 745 (91) | 114 (98) | 631 (89) | 0.0019 |
| Squamous | 62 (8) | 1 (1) | 61 (9) | |
| Other | 15 (2) | 1 (1) | 14 (2) | |
| Missing | 2 | 1 | 1 | |
|
| ||||
| Radiation therapy modality, n (%) | ||||
| 3D | 349 (42) | 49 (42) | 300 (42) | <0.0001 |
| IMRT | 349 (42) | 67 (57) | 282 (40) | |
| Proton | 126 (15) | 1 (1) | 125 (18) | |
|
| ||||
| Induction chemotherapy, n (%) | ||||
| Yes | 369 (45) | 77 (66) | 292 (41) | <0.0001c |
| No | 455 (55) | 40 (34) | 415 (59) | |
|
| ||||
| Type of concurrent chemotherapy, n (%) | ||||
| FP | 253 (31) | 33 (28) | 220 (31) | 0.0116 |
| FT | 422 (52) | 51 (44) | 371 (53) | |
| Other | 143 (17) | 32 (28) | 111 (16) | |
| Missing | 6 | 1 | 5 | |
|
| ||||
| Type of surgery, n (%) (surgery only) | ||||
| Trans-thoracic | 2 (0.5) | 2 (3) | 0 | <0.0001 |
| Ivor-Lewis | 312 (76) | 27 (47) | 285 (81) | |
| Other | 96 (23) | 29 (50) | 67 (19) | |
Generalized Fisher's exact test.
Wilcoxon rank sum test.
Fisher's exact test.
Table 2. Association between radiation volume and prognostic covariates in the 222 matched patients.
| Type of Radiation Volume | ||||
|---|---|---|---|---|
|
|
||||
| Measure | All Pairs (N=222) | Extended Locoregional (N=111) | Locoregional (N=111) | Unadjusted P-Valuea |
| Age at diagnosis (years) | ||||
| Mean | 61.3 | 61.4 | 61.1 | 0.91b |
| Minimum, Maximum | 21.0, 87.0 | 21.0, 85.0 | 22.0, 87.0 | |
|
| ||||
| Gender, n (%) | ||||
| Male | 181 (82) | 88 (79) | 93 (84) | 0.49c |
|
| ||||
| Tumor stage, n (%) | ||||
| 1 | 3 (1) | 1 (1) | 2 (2) | 0.69 |
| 2 | 53 (24) | 25 (23) | 28 (25) | |
| 3 | 151 (68) | 79 (71) | 72 (65) | |
| 4a | 15 (7) | 6 (5) | 9 (8) | |
|
| ||||
| Tumor length (cm) | ||||
| n | 182 | 86 | 96 | 0.62b |
| Mean | 5.5 | 5.6 | 5.4 | |
| Minimum, Maximum | 1.0, 15.0 | 1.0, 13.0 | 1.0, 15.0 | |
|
| ||||
| Karnofsky performance status | ||||
| Mean | 85.9 | 85.7 | 86.2 | 0.53b |
| Minimum, Maximum | 60.0, 100.0 | 60.0, 100.0 | 70.0, 100.0 | |
|
| ||||
| Number of comorbidities | ||||
| Mean | 1.1 | 1.1 | 1.1 | 0.44b |
| Minimum, Maximum | 0, 5.0 | 0, 4.0 | 0, 5.0 | |
|
| ||||
| Tumor differentiation, n (%) | ||||
| Well differentiated | 1 (0.5) | 1 (1) | 0 | 0.78 |
| Moderate | 78 (35) | 40 (36) | 38 (35) | |
| Poor | 141 (64) | 69 (63) | 72 (65) | |
| Missing | 2 | 1 | 1 | |
|
| ||||
| Tumor histology, n (%) | ||||
| Adeno | 220 (99) | 110 (99) | 110 (99) | 1 |
| Squamous | 2 (1) | 1 (1) | 1 (1) | |
| Other | 0 | 0 | 0 | |
|
| ||||
| Radiation therapy modality, n (%) | ||||
| 3D | 97 (44) | 46 (41) | 51 (46) | 0.34 |
| IMRT | 123 (55) | 65 (59) | 58 (52) | |
| Proton | 2 (1) | 0 | 2 (2) | |
|
| ||||
| Induction chemotherapy, n (%) | ||||
| Yes | 135 (61) | 73 (66) | 62 (56) | 0.17c |
| No | 87 (39) | 38 (34) | 49 (44) | |
|
| ||||
| Type of concurrent chemotherapy, n (%) | ||||
| FP | 56 (25) | 31 (28) | 25 (23) | 0.41 |
| FT | 112 (50) | 51 (46) | 61 (55) | |
| Other | 54 (24) | 29 (26) | 25 (23) | |
|
| ||||
| Type of surgery, n (%) | ||||
| None | 112 (50) | 56 (50) | 56 (50) | 1.00 |
| Ivor-Lewis | 54 (24) | 27 (24) | 27 (24) | |
| Other | 56 (25) | 28 (25) | 28 (25) | |
Generalized Fisher's exact test.
Wilcoxon rank sum test.
Fisher's exact test.
Impact of radiation volumes on acute toxicity during chemoradiation
Please refer to Table 3. Compared to locoregional volumes, treatment with extended locoregional radiation resulted in higher rates of grade 3 anorexia (adjusted p=0.0178), grade 3 dysphagia (adjusted p=0.0010), and grade 3 esophagitis (adjusted p=0.0511). There was also increased odds of acute grade 2+ GI toxicity, defined as a combination of weight loss, nausea/vomiting, fatigue, and/or anorexia, among patients treated with extended locoregional vs. locoregional radiation (OR=2.92, adjusted p=0.0447). Additionally, 5 vs. 0 patients treated with extended locoregional vs. locoregional radiation, respectively, experienced a protracted radiation course with prolonged number of days on treatment (unadjusted p=0.0313, adjusted p=1.00).
Table 3. Acute CRT-associated toxicity among 111 matched-pair patients treated with locoregional vs. extended locoregional RT.
| Grade, n (%) | |||||
|---|---|---|---|---|---|
|
|
|||||
| Toxicity | 0 | 1-2 | 3 | P-Valuea | Adjusted P-Valued |
| Weight loss | |||||
| Locoregional RT (n=111) | 54 (48) | 55 (50) | 2 (2) | 0.62 | 1.00 |
| Extended Locoregional RT (n=108) | 58 (54) | 49 (45) | 1 (1) | ||
| Nausea/vomiting | |||||
| Locoregional RT (n=111) | 37 (33) | 66 (60) | 8 (7) | 0.0078 | 0.38 |
| Extended Locoregional RT (n=111) | 20 (18) | 72 (65) | 19 (17) | ||
| Fatigue | |||||
| Locoregional RT (n=111) | 41 (37) | 65 (59) | 5 (4) | 0.74 | 1.00 |
| Extended Locoregional RT (n=111) | 40 (36) | 63 (57) | 8 (7) | ||
| Anorexia | |||||
| Locoregional RT (n=111) | 68 (61) | 42 (38) | 1 (1) | 0.0003 | 0.0178 |
| Extended Locoregional RT (n=111) | 47 (42) | 51 (46) | 13 (12) | ||
| Dysphagia | |||||
| Locoregional RT (n=111) | 23 (21) | 82 (74) | 6 (5) | <0.0001 | 0.0010 |
| Extended Locoregional RT (n=111) | 49 (44) | 48 (43) | 14 (13) | ||
| Esophagitis | |||||
| Locoregional RT (n=111) | 23 (21) | 79 (71) | 9 (8) | 0.0010 | 0.0511 |
| Extended Locoregional RT (n=111) | 45 (41) | 52 (47) | 14 (12) | ||
| Dermatitis | |||||
| Locoregional RT (n=111) | 67 (60) | 44 (40) | 0 (0) | 0.0022c | 0.11 |
| Extended Locoregional RT (n=111) | 89 (80) | 22 (20) | 0 (0) | ||
|
| |||||
| OR | 95% CI | P-Valueb | Adjusted P-Valued | ||
|
| |||||
| Acute GI toxicity Grade 2+ | 2.92 | 1.56, 5.49 | 0.0008 | 0.0447 | |
| Extended Locoregional RT vs. Locoregional RT | |||||
|
| |||||
| No, n (%) | Yes, n (%) | P-Valuec | Adjusted P-Valued | ||
|
| |||||
| Finished RT | |||||
| Locoregional RT (n=111) | 4 (4) | 107 (96) | 1.00 | 1.00 | |
| Extended Locoregional RT (n=111) | 4 (4) | 107 (96) | |||
| Prolonged RT Course* | |||||
| Locoregional RT (n=111) | 111 (100) | 0 (0) | 0.0313 | 1.00 | |
| Extended Locoregional RT (n=111) | 106 (95) | 5 (5) | |||
Abbreviations: CI = confidence interval. CRT = chemoradiation. OR = odds ratio. RT = radiation.
Generalized Fisher's exact test.
Fitted stratified logistic regression.
McNemar mid-p test.
p-values adjusted using Hochberg method.
Prolonged RT course is defined as 2+ standard deviations longer than the mean elapsed days between start and end of RT treatment.
Since the location of nodal metastasis largely depends on the location of the primary GEJ tumor, the volume of coverage may need to include a more extensive area in patients with more gastric-centric tumors. Thus, to normalize the nodal risk between the two groups, we repeated the analyses among the subset of 104 patients (52 and 52 patients with locoregional and extended locoregional radiation, respectively) with Siewert type II tumors. While there continued to be a trend toward increased odds of acute grade 2+ GI toxicity with extended locoregional vs. locoregional radiation (OR 2.00, un-adjusted p=0.13, adjusted p=1.00), these results were not statistically significant.
Impact of radiation volumes on postoperative complications
As shown in supplemental Table 1, there was a high association between type of surgery performed and locoregional vs. extended locoregional radiation volume. While patients treated with extended locoregional radiation were more likely to undergo gastrectomies, patients treated with locoregional radiation were more likely to undergo Ivor-Lewis and other esophagectomy approaches. Because the surgical approach commonly differed between patients treated in the two groups, comparison of postoperative complications required controlling for surgical differences between the two.
To account for these surgical differences, we matched by the exact type of surgery before assessing postoperative complication rates between the two (locoregional vs. extended locoregional radiation) groups. Because the only type of surgery that overlapped among those patients who received locoregional vs. extended locoregional radiation was Ivor-Lewis esophagectomy (supplemental Table 1), we could only analyze postop complications among those patients who underwent Ivor-Lewis surgery (N=27 matched pairs). In this subset analysis, after adjusting for multiple testing, locoregional vs. extended locoregional radiation had no impact on the rate of postoperative GI, pulmonary, cardiac, wound complications, or length of hospital stay (p>0.05, Table 4).
Table 4. Postoperative complications among matched pair patients treated with extended locoregional vs. locoregional RT.
| POC among 27 matched pairs treated with Ivor-Lewis surgery only | ||||
|---|---|---|---|---|
|
| ||||
| Postoperative complication | Extended Locoregional RT | Locoregional RT | P-Valuea | Adjusted P-Valuec |
| GI, n (%) | ||||
| Yes | 9 (33) | 4 (15) | 0.15 | 1.00 |
| No | 18 (67) | 23 (85) | ||
| Pulmonary, n (%) | ||||
| Yes | 14 (52) | 10 (37) | 0.30 | 1.00 |
| No | 13 (48) | 17 (63) | ||
| Cardiac, n (%) | ||||
| Yes | 6 (22) | 4 (15) | 0.51 | 1.00 |
| No | 21 (78) | 23 (85) | ||
| Wound/Infection, n (%) | ||||
| Yes | 8 (31) | 1 (4) | 0.0039 | 0.20 |
| No | 18 (69) | 26 (96) | ||
| Unknown | 1 | 0 | ||
| Length of hospital stay (days) | ||||
| n | 26 | 27 | 0.0163b | 0.75 |
| Mean | 18.8 | 13.5 | ||
| Median | 15 | 11 | ||
| Minimum, Maximum | 8.0, 76.0 | 7.0, 47.0 | ||
McNemar mid-p test.
Wilcoxon rank sum test.
p-values adjusted using Hochberg method.
Abbreviations: POC = postoperative complications. RT = radiation.
Among patients with Siewert type II tumors who underwent Ivor-Lewis surgery only (N=16 matched pairs), again there was no difference between locoregional vs. extended locoregional radiation in the rate of postop GI (unadjusted p=0.18, adjusted p=1.00), pulmonary (unadjusted p=0.45, adjusted p=1.00), cardiac (unadjusted p=0.69, adjusted p=1.00), wound/infection complications (unadjusted p=0.0313, adjusted p=1.00), or length of hospital stay (unadjusted p=0.0214, adjusted p=0.97). Data are not shown.
Impact of radiation volumes on patterns of failure
Among the patients who underwent surgery (N=55 matched pairs), locoregional vs. extended locoregional radiation did not affect the pathologic T stage (unadjusted p=0.48, adjusted p=1.00), frequency of positive regional nodes on surgical specimen (36% vs. 38%, unadjusted p=0.83, adjusted p=1.00), or rate of positive perigastric nodes on surgical specimen (15% vs. 22%, unadjusted p=0.38, adjusted p=1.00). This remained true for the subset of patients (N=16 matched pairs) with Siewert type II tumors.
Since one major difference between the radiation volumes is elective coverage of celiac nodes, we analyzed differences in celiac relapse between the two. Patients who had celiac node involvement at diagnosis (n=8) and those with missing data (n=8) were excluded from analysis. Among the 203 remaining patients, 28% and 100% of the patients treated with locoregional vs. extended locoregional radiation, respectively, received elective celiac radiation. Patients treated with locoregional radiation (n=111) were more likely to receive elective celiac coverage if they had node positive or overall clinical stage 3+ disease, although these results were not statistically significant after adjusting for multiple testing (supplemental Table 2). Of the 72 patients who did not receive elective celiac radiation, 2 (2.8%) patients relapsed in the celiac nodes. Of the 124 patients who received elective celiac radiation, 2 patients (1.6%) also relapsed in the celiac nodes. In 3 of these patients, the celiac failures represented the first site of recurrence and occurred simultaneously with distant metastases (supplemental Table 3). No patients failed in the splenic or porta nodes. Of note, the mean dose to the splenic hilum for those who received extended locoregional (n=111) vs. locoregional RT (n=65 with evaluable radiation plans) was 45 Gy vs. 19 Gy (median 20 Gy, range 5-40 Gy), respectively.
Impact of radiation volumes on survival outcomes
Based on stratified Cox proportional hazards regression models on the 111 matched pairs, there was no association between locoregional vs. extended locoregional volumes and OS, distant PFS, or locoregional PFS among the subset of patients who received surgery, those who received definitive chemoradiation alone, and those with Siewert type 2 tumors (Table 5).
Table 5. Fitted conditional Cox models assessing effect of extended locoregional vs. locoregional RT on survival outcomes.
| Survival outcomes among all extended locoregional vs. locoregional RT matched pairs | |||||
|---|---|---|---|---|---|
| Outcome | # matched pairs | HR | 95% CI | P-Value | Adjusted P-Value* |
| Overall survival | 109 | 0.97 | 0.62, 1.54 | 0.91 | 1.00 |
| PFS – distant metastasis | 86 | 1.1 | 0.67, 1.78 | 0.71 | 1.00 |
| PFS – local regional control | 71 | 0.86 | 0.50, 1.48 | 0.58 | 1.00 |
| Survival outcomes among extended locoregional vs. locoregional RT matched pairs treated with surgery | |||||
| Outcome | # matched pairs | HR | 95% CI | P-Value | Adjusted P-Value* |
| Overall survival | 54 | 0.78 | 0.39, 1.56 | 0.48 | 1.00 |
| PFS – distant metastasis | 42 | 0.93 | 0.45, 1.93 | 0.85 | 1.00 |
| PFS – local regional control | 39 | 0.67 | 0.30, 1.48 | 0.32 | 1.00 |
| Survival outcomes among extended locoregional vs. locoregional RT matched pairs treated with definitive CRT | |||||
| Outcome | # matched pairs | HR | 95% CI | P-Value | Adjusted P-Value* |
| Overall survival | 55 | 1.16 | 0.63, 2.14 | 0.64 | 1.00 |
| PFS – distant metastasis | 44 | 1.25 | 0.65, 2.41 | 0.51 | 1.00 |
| PFS – local regional control | 32 | 1.08 | 0.51, 2.29 | 0.85 | 1.00 |
P-values adjusted using the Hochberg method.
Abbreviations: HR = hazard ratio; CI = confidence interval; PFS = progression-free survival.
Discussion
While there is growing evidence supporting the effectiveness of chemoradiation for the treatment of non-metastatic esophageal cancer1-4, details regarding target volume delineation and radiation fields have not been addressed in randomized-controlled trials. This has led to variable radiation volumes and elective nodal coverage used in recent trials2, 4, 13. Thus, the goal of our current study was to assess whether additional coverage of regional splenic/porta nodes is necessary for the treatment of patients with GEJ tumors. Using matched pair analyses and adjustment for multiple testing, we found no differences in patterns of failure or survival between the two RT volumes, although acute GI toxicity was higher among patients treated with extended locoregional radiation. Extended locoegional RT also resulted in more grade 3 dysphagia/esophagitis despite less radiation dose; this may be related to more coverage of the stomach/liver and worse vomiting associated with extended locoregional radiation leading to more reflux/acid-damage to the esophagus.
One of the important goals of chemoradiation is to reduce the primary tumor volume without increasing toxicity. However, achieving this goal hinges on defining the appropriate radiation volume. One obstacle to accurately defining the target for GEJ tumors include the controversy of whether GEJ tumors should be classified as gastric or esophageal tumors. The UICC TNM 6th edition classified GEJ tumors as gastric cancers14; but, the recent AJCC TNM classification changed the categorization of GEJ tumors, such that all tumors located within 5cm below the GEJ and infiltrating the junction are now classified as esophageal cancers. Thus, all Siewert types I, II, and III GEJ tumors are considered and staged as esophageal cancers.
This debate on which Siewert type of GEJ carcinomas should be classified as esophageal vs. gastric cancers is important because patterns of spread and failure differ between esophageal vs. gastric primaries. Although NCCN recommends Siewert type I and II tumors to be generally managed with radiation guidelines applicable to esophageal cancers and type III tumors to be managed with guidelines applicable to either esophageal or gastric cancers, these represent category 2A recommendations with lower level of evidence. Overall, it is probably most accepted that type I GEJ tumors should be treated as distal esophageal carcinomas, but how type II and III GEJ tumors should be treated are more controversial9, 15. The EORTC expert opinion14 and another pathological review16 of the location and frequency of recurrences suggested elective nodal coverage of para-esophageal, peri-gastric, and celiac nodes for type I GEJ tumors, but differ from NCCN and our study results by recommending additional splenic nodal coverage for type II/III GEJ tumors. However, those recommendations were based on out-dated patterns of failure data from patients not treated with more modern techniques and often do not account for possible eradiaction of microscopic nodal disease with neadjuvant chemoradiation or subsequent surgery. Moreover, the recommendations were not based on studies that directly compared whether difference in elective nodal coverage with radiation would impact patterns of failure.
To more clearly define a radiation target volume for GEJ tumors, our study compared locoregional vs. extended locoregional radiation volumes in terms of toxicity and clinical outcomes. Because locoregional volumes resulted in less acute chemoradiation-associated GI toxicity without compromising relapse/survival, our results suggest that coverage of first-echelon and celiac nodes alone may be adequate for Siewert type I/II GEJ tumors. We cannot draw definitive conclusions on whether type III tumors should be treated with locoregional vs. extended locoregional radiation, since only 1 patient with type III tumor was treated with locoregional radiation. With respect to whether the celiac axis should receive elective coverage for GEJ carcinomas, while we confirm previously published findings17 in that celiac relapse was rare regardless of whether the patient received elective celiac radiation, there may have been selection bias in determining who received elective celiac coverage among those treated with locoregional radiation (supplemental Table 2). Other potential reasons for why some but not all patients were chosen for elective celiac coverage among those who received locoregional radiation cannot be fully elucidated. This selection bias, in combination with the lack of effective salvage treatments for celiac failures, would caution against omitting prophylactic celiac treatment in GEJ tumors. Given that no patients failed in the porta or splenic nodes, however, our results do suggest that omission of porta/splenic regional nodal coverage may be safe in type I-II GEJ tumors. While we acknowledge that patients who did not receive elective splenic coverage still got an incidental low scatter dose to the splenic hilum, which may or may not have contributed to disease control, our study still suggests that treating splenic nodes to high dose is not necessary given comparable disease control and reduced toxicity with omission of elective coverage.
Our study is subject to limitations intrinsic to a retrospective dataset. First, patients were not randomized between radiation fields and consequently, correction of many potential biases was needed. While we were able to obtain closely matched pairs to correct for biases of known variables, possible effects of additional external confounding factors cannot be excluded. Initially, we considered alternative methods of correcting for bias, including stratification and inverse probability of treatment weights (IPTW) analysis. However, there were far too many strata to analyze feasibly; and, IPTW was not possible due to high collinearity among many prognostic covariates and other treatment variables, which made estimation of propensity scores unreliable. Second, radiation doses and techniques were not standardized. Third, physician and surgeon treatment bias is inherent in any retrospective analysis. Lastly, the relatively small sample sizes used in the subset analyses limit the reliability of any conclusions.
Nevertheless, since a randomized trial to compare radiation field coverage based on Siewert grouping is unlikely to be conducted, our study offers a limited but worthwhile attempt to delineate the ideal target volume for GEJ tumors. In conclusion, our data suggest that selected type I and II non-metastatic GEJ tumors can be treated sufficiently with elective coverage of adjacent paraesophaegal (RTOG nodal station 8), perigastric (RTOG nodal station 15-17), and celiac lymph nodes (RTOG nodal station 20); elective inclusion of more regional splenic/porta nodes (RTOG nodal stations 19/18) can be safely omitted from the CTV. Whether type III GEJ tumors can be treated with locoregional or more conventional extended locoregional radiation volumes needs further investigation.
Supplementary Material
Acknowledgments
Source of Funding: Funding was provided in part by The University of Texas MD Anderson Cancer Center and by the National Cancer Institute Cancer Center Support Grant CA016672.
Footnotes
Conflicts of Interest: The authors have no conflicts of interest to report.
References
- 1.Stahl M, Walz MK, Stuschke M, et al. Phase III comparison of preoperative chemotherapy compared with chemoradiotherapy in patients with locally advanced adenocarcinoma of the esophagogastric junction. Journal of clinical oncology : official journal of the American Society of Clinical Oncology. 2009;27:851–856. doi: 10.1200/JCO.2008.17.0506. [DOI] [PubMed] [Google Scholar]
- 2.Tepper J, Krasna MJ, Niedzwiecki D, et al. Phase III trial of trimodality therapy with cisplatin, fluorouracil, radiotherapy, and surgery compared with surgery alone for esophageal cancer: CALGB 9781. Journal of clinical oncology : official journal of the American Society of Clinical Oncology. 2008;26:1086–1092. doi: 10.1200/JCO.2007.12.9593. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Urba SG, Orringer MB, Turrisi A, et al. Randomized trial of preoperative chemoradiation versus surgery alone in patients with locoregional esophageal carcinoma. Journal of clinical oncology : official journal of the American Society of Clinical Oncology. 2001;19:305–313. doi: 10.1200/JCO.2001.19.2.305. [DOI] [PubMed] [Google Scholar]
- 4.van Hagen P, Hulshof MC, van Lanschot JJ, et al. Preoperative chemoradiotherapy for esophageal or junctional cancer. The New England journal of medicine. 2012;366:2074–2084. doi: 10.1056/NEJMoa1112088. [DOI] [PubMed] [Google Scholar]
- 5.Siewert JR, Holscher AH, Becker K, et al. Cardia cancer: attempt at a therapeutically relevant classification. Der Chirurg; Zeitschrift fur alle Gebiete der operativen Medizen. 1987;58:25–32. [PubMed] [Google Scholar]
- 6.Wijnhoven BP, Siersema PD, Hop WC, et al. Adenocarcinomas of the distal oesophagus and gastric cardia are one clinical entity. Rotterdam Oesophageal Tumour Study Group. The British journal of surgery. 1999;86:529–535. doi: 10.1046/j.1365-2168.1999.01082.x. [DOI] [PubMed] [Google Scholar]
- 7.Huang Q, Shi J, Feng A, et al. Gastric cardiac carcinomas involving the esophagus are more adequately staged as gastric cancers by the 7th edition of the American Joint Commission on Cancer Staging System. Modern pathology : an official journal of the United States and Canadian Academy of Pathology, Inc. 2011;24:138–146. doi: 10.1038/modpathol.2010.183. [DOI] [PubMed] [Google Scholar]
- 8.Kim HI, Cheong JH, Song KJ, et al. Staging of adenocarcinoma of the esophagogastric junction: comparison of AJCC 6th and 7th gastric and 7th esophageal staging systems. Ann Surg Oncol. 2013;20:2713–2720. doi: 10.1245/s10434-013-2898-5. [DOI] [PubMed] [Google Scholar]
- 9.Suh YS, Han DS, Kong SH, et al. Should adenocarcinoma of the esophagogastric junction be classified as esophageal cancer? A comparative analysis according to the seventh AJCC TNM classification. Annals of surgery. 2012;255:908–915. doi: 10.1097/SLA.0b013e31824beb95. [DOI] [PubMed] [Google Scholar]
- 10.Edge SB, B D, Compton CC, et al. AJCC cancer staging manual. New York: Springer; 2010. [Google Scholar]
- 11.Rosenbaum PR, Rubin DB. Constructing a Control-Group Using Multivariate Matched Sampling Methods That Incorporate the Propensity Score. American Statistician. 1985;39:33–38. [Google Scholar]
- 12.Ho DE, Imai K, King G, et al. Matching as nonparametric preprocessing for reducing model dependence in parametric causal inference. Political Analysis. 2007;15:199–236. [Google Scholar]
- 13.Swisher SG, Winter KA, Komaki RU, et al. A Phase II study of a paclitaxel-based chemoradiation regimen with selective surgical salvage for resectable locoregionally advanced esophageal cancer: initial reporting of RTOG 0246. International journal of radiation oncology, biology, physics. 2012;82:1967–1972. doi: 10.1016/j.ijrobp.2011.01.043. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Matzinger O, Gerber E, Bernstein Z, et al. EORTC-ROG expert opinion: radiotherapy volume and treatment guidelines for neoadjuvant radiation of adenocarcinomas of the gastroesophageal junction and the stomach. Radiotherapy and oncology : journal of the European Society for Therapeutic Radiology and Oncology. 2009;92:164–175. doi: 10.1016/j.radonc.2009.03.018. [DOI] [PubMed] [Google Scholar]
- 15.Pedrazzani C. Should Adenocarcinoma of the Esophagogastric Junction Be Classified as Gastric or Esophageal Cancer, or Else as a Distinct Clinical Entity? Annals of surgery. 2014 doi: 10.1097/SLA.0000000000000524. [DOI] [PubMed] [Google Scholar]
- 16.Meier I, Merkel S, Papadopoulos T, et al. Adenocarcinoma of the esophagogastric junction: the pattern of metastatic lymph node dissemination as a rationale for elective lymphatic target volume definition. International journal of radiation oncology, biology, physics. 2008;70:1408–1417. doi: 10.1016/j.ijrobp.2007.08.053. [DOI] [PubMed] [Google Scholar]
- 17.Amini A, Xiao L, Allen PK, et al. Celiac node failure patterns after definitive chemoradiation for esophageal cancer in the modern era. International journal of radiation oncology, biology, physics. 2012;83:e231–239. doi: 10.1016/j.ijrobp.2011.12.061. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.