Assessing failure patterns of radical intent radiation strategies in patients with locally advanced carcinoma of the esophagus

Abstract Background Patterns of failure following definitive CRT (dCRT) are different as compared to neoadjuvant chemoradiotherapy (NACRT) with increased locoregional failures documented with dCRT. Aim To document failure patterns in patients with esophageal carcinoma treated with neoadjuvant and definitive intent radiation strategies. Methods Subjects were 123 patients treated with two chemoradiotherapy strategies. Group 1 (n = 99) underwent dose escalated definitive chemoradiotherapy (dCRT), Group 2 (n = 24) received neoadjuvant chemoradiotherapy (NACRT) followed by surgery. Cumulative incidence of locoregional failure (LRF), local failure (LF), regional lymph node failure (RLNF), and distant metastasis (DM) were computed; differences between the groups was evaluated using log rank test. Univariable and multivariable predictors of failure were identified using Cox regression analysis. Results Cumulative LRF: 64% in Group 1 vs 35% in Group 2 (P = .050). Cumulative LF: 59% in Group 1 vs 12% in Group 2 (P = .000). Cumulative RLNF: 30% in Group 1 vs 24% in Group 2 (P = .592). Most common RLNF: mediastinum for both groups (6% vs 12.5%, respectively). Distant metastasis: 40.4% Group 1 vs 17% Group 2 (P = .129), predominantly lung (Group 1, 5%), and nonregional nodes (Group 2, 8.3%). Univariate analysis identified age ≤50, absence of concurrent chemotherapy, dose ≤50 Gy, and incomplete radiotherapy to predict higher odds of LRF and DM for Group 1; absence of comorbidities predicted for lower odds of LRF for Group 2. Age ≤50 predicted for higher odds of RNLR for Group 1, while absence of comorbidities predicted for lower odds of RNLR in Group 2. Multivariate analysis identified age ≤50, incomplete radiotherapy, and absence of concurrent chemotherapy to predict higher odds of LRF for Group 1. Age ≤50, absence of concurrent chemotherapy predicted higher odds of DM for Group 1. Absence of comorbidity predicted lower odds of LRF in Group 2. Conclusion LRF is common in both groups, with LF being predominant in dCRT as opposed to RNLF in NACRT. Age ≤50, absence of concurrent chemotherapy is a predictor of LRF and DM in dCRT.

Conclusion: LRF is common in both groups, with LF being predominant in dCRT as opposed to RNLF in NACRT. Age ≤50, absence of concurrent chemotherapy is a predictor of LRF and DM in dCRT. Cancer Network guidelines recommend that CTV should include the areas at risk for microscopic disease as well as elective nodal regions, depending on the location of the primary tumor in the esophagus. 5 The choice of elective nodal regions depends on the probability of lymph node involvement as understood from patterns of failure and surgical series. [6][7][8] However, still there are no consistent standards for constructing proper CTV worldwide. 9 For both dCRT and NACRT, similar principles of tumor volume delineation are applied. Different methods and opinions have been reported and practiced in various countries. 10 Therefore, understanding relapse patterns is important, not only to gain insight into the effectiveness of the combined treatment but also to foster improvements in designing prophylactic target volumes.
The other unresolved question in esophageal cancers is the dose in the dCRT setting. Although radiation dose escalation has failed to improve local control or survival, a dose of 60 Gy is more popular in Asian countries 11 where SCC is the predominant histological type; with modern radiotherapy (RT) techniques, the role of dose escalation is being revisited in randomized trials. 12 This study attempts to discern the failure patterns in our population of predominantly thoracic esophageal SCC (ESCC) treated after two radical therapeutic approaches, namely escalated dose of dCRT and NACRT followed by surgery, and to help optimize the field design with particular reference to designing nodal CTV for both these approaches. We assessed a variety of patients from the perspective of their disease-related and treatment characteristics for their potential utility as predictors of failure, reasoning that such factors may uncover opportunities for targeted intensification of therapy in patients with poor prognostic factors.

| MATERIAL AND METHODS
Between January 2011 and December 2014, a total of 123 consecutively registered patients ( Figure S1)  Group 2, Neoadjuvant CRT (NACRT) followed by Surgery: operable and fit patients, T2-4 ± node-positive with performance ≥80, were selected for this approach. These patients received a dose of 45 Gy in 25 fractions with concurrent weekly Cisplatin 35 mg/m 2 . They were reassessed both clinically and radiologically with CECT scan after NACRT in a multidisciplinary clinic for surgery. The majority underwent a transthoracic resection with two-field lymph node dissection.

| Radiotherapy target volumes
The gross tumor volume was defined by combining information from computed tomography, esophagogastroduodenoscopy, barium scan, and included all involved local and regional lymph nodes ( Figure S2).
The radiotherapy clinical target volume was defined by the gross tumor volume plus a minimum of 1 cm radially and 3 cm longitudinally to a dose of 36 Gy, followed by a cone down (2 cm longitudinal margins to gross tumor volume) to a dose of 45 Gy or 60 to 66 Gy. Gross nodal disease identified as enlarged nodes on CECT scans received a dose of 45 Gy in Group 2 or 60 Gy in Group 1. In the patients with gastroesophageal junction involvement, and who were not being undertaken for surgery, the dose was restricted to 55.8 Gy in 31 fractions respecting stomach tolerance. Prophylactic nodal radiation to the supraclavicular (SCF) region was used in cases when the bulk of disease was supracarinal (primarily Group 1). Doses to the prophylactic SCF nodal volumes were 45 Gy/25#. Three-dimensional conformal treatment was mostly used with phase I treated by AP/PA fields followed by phase II treated by three fields, both with multileaf collimator conformation.  was mapped as infield if located at a primary site within the planned field, or out of field recurrence if located outside the planned field.
The institutional review board approved this study and waived the requirement for written informed consent due to its retrospective nature.

| Statistics
Demographic data and failure patterns were summarized using crude percentages. The differences between the groups were evaluated   Table 1. Generally, local failure was the prevalent mode of failure in Group 1 (seen in 48.5% of patients) as opposed to nodal failure in Group 2 (seen in 17% of patients; Table 2).

| Loco-regional failure
Locoregional failures were documented in 52.5% (n = 52) patients in Group 1 and 25% (n = 6) patients in group 2 (P = .022; Figure 1A,B; Tables 2-4). Cumulative incidence of LRF in Group 1 vs Group 2 at 5 years was 64% vs 35% (P = .05) as depicted in Figure 1A. Local failures were documented in 48 patients in Group 1 as opposed to 2 patients in Group 2 (48.5% vs 8.3%, P = .000). Local failures were infield in Group 2, and mostly documented as local persistence of disease at primary site in Group 1.
Regional failures were documented in 13 patients in Group 1 as opposed to 4 in Group 2 (13% vs 17%, P = .659). Cumulative incidence of RLNF at 5 years was 30% vs 24% in Group 1 vs 2 (P = .592) as depicted in Figure 1B. The crude incidence of RLNF without distant metastasis were seen in 5% vs 12.5% (five and three patients in Group 1 and Group 2, respectively), P = .356. Most common RLNF in Group 1 was mediastinal (n = 6, 6%) followed by celiac (n = 4, 4%) and SCF (n = 3, 3%). Most common RLNF in Group 2 was mediastinal (n = 3, 12.5%) followed by SCF (n = 1, 4%). Among regional failures, most were out of field in Group 1 (n = 9, 69.2%) located in SCF and celiac regions as compared to Group 2 in which most failures were in field (n = 3, 75%) located in mediastinal regions. Out of field regional failures without distant metastasis were seen in three patients (3/99 = 3%) in Group 1 and one patient (1/24 = 4%) in Group 2. Location of disease had an impact on location of regional failure as depicted in Figure 2. recurrence. Due to low local, regional failure events in Group 2, multivariable analysis was not done (Table 4).

| DISCUSSION
The management of localized or locally advanced esophageal carcinoma is difficult because of the high likelihood of loco-regional recurrences. In this study, we found that locoregional disease was the predominant reason for treatment failure in both groups( Figure 1A), and the crude incidence of LRF was significantly high in Group 1 vs Group 2 (52.5% vs 25%, P = .022). These results were no different from other studies suggesting that the LRF is a common mode of relapse after NACRT or dCRT. [13][14][15] Locoregional failure without distant metastasis was seen in nearly half the patients in Group 1 (49.4%) as opposed to one-fourth in Group 2 (25.4%). As in our study, LRR rates of 13% to 25% 16,17 are reported after NACRT as compared to 43% to 55% 18 in patients receiving dCRT. Isolated local failure without concomitant distant metastasis was seen in 39 and 2 patients (39% vs 8.3%, P = .001). Local failures were common in Group 1 as opposed to regional nodal failures in Group 2. The crude and cumulative incidence of failures is depicted in Table 2.
F I G U R E 2 Location of tumor and location of regional nodal failure (in absolute numbers) for Group 1 (dCRT) and Group 2 (NACRT). For both groups mediastinal failures were common F I G U R E 3 Comparison of cumulative incidence of distant metastasis (DM) between Group 1 (d CRT) and 2 (NACRT). The 5 years cumulative incidence of distant metastasis is 40% vs 17% in Group 1 vs 2 (logrank P = .129) The evidence underlying RT planning volumes, in particular CTV, is poorly defined, and very little published data are available regarding the margins required to allow for microscopic disease extension from the defined gross target volume (GTV). 7,19 In published series for dCRT, the margins have varied from elective treatment of the whole esophagus with the SCF to a defined 3 to 5 cm SI margin above and below the defined GTV. 20 For preoperative CRT, a fixed margin above and below the GTV has been used in trials, 20 but controversy exists regarding the use of prophylactic nodal radiation. 19 The esophageal submucosa has an extensive lymphatic vertical distribution, and numerous studies have suggested that multiple levels and skipped node metastases are commonly observed in esophageal SCC. 21,22 As per our analysis, RLNF rates without distant metastasis were seen in 5% vs 12.3% in Group 1 vs Group 2, P = .356 (five and three patients, respectively, Table 3 Out of field regional failures 9, (2 = SCF, 4 = Celiac, 3 = Mediastinal) 1(SCF)

9% 4%
Out of field node × regional nodal failur Others (multiple sites) 6(6%) 1(4.1%) Note: Infield regional failures: failures located within the radiotherapy portal; Out of Field regional: failures located outside the radiotherapy portal. as studies have clearly shown an increased risk of mortality with extensive lymph node dissection. 30 Regional nodal failures were mostly out of field in the dCRT (n = 9/13, 69%) group as compared to being infield in the NACRT (n = 3/4, 75%) group. Eloubeidi et al 29 reported that longer lesions led to more lymph node metastasis outside the target volume, and as patients with dCRT had a greater length, most failures were out of field. Overall isolated out-of-field regional failure without distant metastasis in the nontreated elective regions was only 3% to 4%, in both the groups, similar to the rates of 3% reported by Button et al,7 and it is unlikely that treating additional lymph node regions would have a meaningful clinical impact as the micrometastases at distant sites would nullify the benefits of clearance of locoregional disease.
Nevertheless, local failures in both the dCRT and NACRT were infield.
These results further support focusing the attention of local therapies  In conclusion, LRF are a frequent occurrence both in dCRT and NACRT with RLNF being common in NACRT as opposed to LF in dCRT. Among regional nodes, mediastinal nodal failure was the most common in both the groups. With respect to field design, regional failures in dCRT were out of field, but most of them had concurrent distant metastasis. Conversely, regional failures in the NACRT group were infield (high mediastinum), and without concurrent distant metastasis. Concurrent chemotherapy enhances locoregional control and decreases distant metastasis in patients undergoing dCRT while young age predicts for increased LRF and DM. Modification of RT fields would unlikely be helpful. Optimization of systemic therapy for dCRT and field of dissection for NACRT might be warranted.

CONFLICT OF INTEREST
The authors have stated explicitly that there are no conflicts of interest in connection with this article.