Delayed esophagectomy for adenocarcinoma is associated with a negative impact on long‐term survival and an increased risk of perioperative morbidity

Delayed esophagectomy (DE) following chemoradiation therapy (CXRT) for esophageal carcinoma is undertaken in selected patients. This study aimed to assess both short‐term outcomes and long‐term survival for patients with adenocarcinoma undergoing DE.


| Variables of analysis
Cohort demographic variables including age, gender, race/ethnicity, insurance status, great circle distance from hospital, urban/ rural setting, facility type, Charlson comorbidity score (CCS), were defined from NCDB database.Income was defined as < or ≥$48 000/year, race was categorized as white/nonwhite.Margin status was defined as positive (R1/2) or negative (R0).AJCC Stage grouping was determined by 7th edition T, N, M, status descriptors.

| Statistical analysis
The patient cohort was divided into two groups based on the timing of surgery following completion of radiation therapy: (1) delayed esophagectomy (DE), ≥90 days, (2) nondelayed esophagectomy (NDE), <90 days.Demographic and tumor-specific variables were compared with the chi-square test for categorical variables and t test for continuous variables.Unadjusted 5-year overall survival (OS) was determined using the Kaplan-Meier log-rank test.Cox proportional hazards regression model was created to identify variables predictive of mortality, including DE. Proportional hazard assumptions were tested by Schoenfeld residuals.Postoperative outcomes including 30-, 90-day mortality, length of stay (LOS), and resection margin status were assessed.Subgroup survival analysis for the DE cohort, grouped by cT and cN status were performed via Kaplan-Meier method.All p values were two-sided unless otherwise specified, a value <0.05 was considered significant.Analyses were performed using SAS 9.3 (SAS Institute).

| Demographics
A total of 8362 patients met inclusion criteria.Of those patients, 7598 (91%) underwent NDE and 764 (9%) undergoing DE.Patients who underwent DE were more likely to be older (>69 years), nonwhite race, and Medicare/Medicaid insurance status (Table 1).No differences were seen with respect to CCS, facility type, or sex.The clinical and pathologic T-and N-status did vary between groups, with NDE patients more likely to be cN > 0, cT > 2 and pN > 0, pT > 2 (Table 2).

| Survival analysis
The unadjusted survival analysis favored NDE when compared to patients undergoing DE.Five-year OS calculated from the date of completion of chemoradiation was 36% versus 30%, (p < 0.01), with a median survival 33.6 versus 27 months (p < 0.01) (Figure 1).Multivariable Cox proportional hazards regression modeling was used to identify independent factors associated with mortality.
Results demonstrated DE as an independent risk factor for mortality (hazard ratio 1.11; confidence interval 1.003-1.23,p = 0.044).Increasing age, male sex, advancing cT-status, and cN > 0 were also associated with an increased risk of death (Table 3).When cN > 0 patients alone were evaluated, no survival difference was seen between DE and NDE, with 5-year OS of 31% and 33%, respectively (p = 0.21) (Figure 2).

| DE subgroup
Subgroup survival analyses were performed on patients undergoing DE to identify tumor/clinical characteristics predictive of mortality.
Multivariable Cox Regression identified treatment at nonacademic/ university medical centers as being associated with an increased risk of mortality (Table S1), while cT-and cN-status were not predictive.

| Timing of esophagectomy
The number of days between esophagectomy and the date of completion of radiation therapy was calculated; 67% of patients underwent surgery between 0 and 59, 25% of patients between 60 and 89, and 9% of patients 90 days or more after completion of last session of radiation therapy.The median number of days between surgery and radiation therapy completion for patients in the DE group was 108 days, compared to 50 days in the NDE group (Table 5).Five-year OS survival was worse for the >90-day DE group (Figure 3).

| Regional radiation dosage
The median dosage of regional radiation was 5000 cGy in the DE group, compared to 4500 cGy in the NDE group.The lower quartile to upper quartile range for radiation dosage was the same between the DE and NDE groups, at 4500-5040 cGy (Table S2).

| Perioperative outcomes
Perioperative outcomes were evaluated comparing NDE to DE.The risk of postoperative mortality at 30 days was 2.8% for NDE versus 4.7% for DE (p < 0.01).Equivalent findings were seen at 90 days; 6.6% versus 11.4% (p < 0.0001).When readmission was assessed, there was no significant difference between NDE and DE.The inpatient median LOS was increased by 1 day for patients undergoing DE compared to NDE (10 vs. 9 days, p < 0.001).Operative pathologic outcomes also favored NDE over DE, with positive surgical margins (>R0) seen in 4.5% versus 7.5% (p < 0.001) (Table 6).

| DISCUSSION
The current guidelines for treatment of local/regionally advanced esophageal adenocarcinoma recommend trimodality therapy, consisting of neoadjuvant chemoradiotherapy followed by definitive surgical resection.Although trimodality therapy has been shown to improve OS, the concept of DE evolved out of the necessity to F I G U R E 1 Kaplan-Meier overall survival curve of delayed esophagectomy versus nondelayed esophagectomy patients.
provide "definitive" therapy to patients initially deemed medically unfit for surgery. 3The utility of definitive nonsurgical treatment was extrapolated to include patients who would otherwise be considered medically suitable candidates for resection.The utility of an active surveillance strategy with DE is inferred from randomized studies demonstrating pCR rates of up to 33% following CXRT. 2,3,8In appropriately selected patients demonstrating clinical complete response after neoadjuvant CXRT, active surveillance to postpone surgical resection may represent a viable organ-sparing treatment strategy. 9In the case of early-stage disease, esophagus-sparing approaches have been successfully applied in the form of endoscopic ablation and/or resectional techniques.Whether a similar approach can be applied to patients with more advanced disease may represent a bridge too far.To date, limited retrospective studies have provided mixed results with respect to short and midterm outcomes for intentional DE following CXRT. 10We sought to assess outcomes for esophageal adenocarcinoma based on the time frame from completion of CXRT to definitive resection.
A comprehensive therapy for esophageal cancer necessitates management of failure risk for both local-regional as well as distant disease.Systemic chemotherapy provides effective reduction in the risk of mortality and disease recurrence when administered either alone or as combined CXRT. 2,11For advanced primary tumors, trimodality patients received two separate local therapies in the form of radiotherapy (RT) and surgical esophagectomy; eliminating surgical resection is predicated on the assumption that RT alone is adequate for local disease control. 12,13For those cases with presumed pCR, intentional DE may be considered at the time of a local-only recurrence when identified on surveillance.When evaluating the pattern of recurrence for patients completing trimodality therapy, the frequency of isolated local-regional failure is as low as 10%, while the majority of recurrences occur distant or as combined localregional and distant; in a prior series evaluating the efficacy of intensive surveillance, when local-regional recurrence is identified in the context of scheduled surveillance, intervention may provide little survival benefit. 14In our unadjusted survival analysis, delayed definitive surgical intervention was associated with a significant reduction in 5-year overall and median survival.Additionally, Cox regression identified DE as being an independent predictor of mortality.Attempting intervention in the form of DE at the time of recurrence following definitive CXRT or following prolonged recovery after neoadjuvant therapy may be selecting those unlikely perceive benefit and subject them to a higher-risk intervention.
Our cohort was further stratified based on the number of days elapsed between the completion of radiation therapy and the date of surgery: patients who underwent surgery under 60 days after completion of RT, between 60 and 89 days after RT completion, and 90 days or more between RT completion and surgery.This comparative analysis illustrated a decrease in 5-year OS as the time between RT completion and surgery increased.
In our series, both the 30-and 90-day mortality were nearly doubled for DE, as compared to NDE.This was in addition to an increase in the margin positive resection rates from 4.2% to 7.1%.
These results appear consistent with other recently published series evaluating perioperative outcomes following DE.In the series from anastomotic leak rate was 17.2% as compared to 10.7% in the NDE group. 16Potential treatment-related complications that might explain these differences include CXRT-associated morbidity as the cause of delay in the DE group, resulting in suboptimal performance status at the time of esophagectomy.The perioperative risk profile of DE suggests judicious patient selection, as the risk of adverse events or poor surgical pathologic outcomes (margin positive) is significant.
Extrapolating from studies of patients treated with definitive nonsurgical modalities, the RTOG 85-01 randomized trial and INT 0123 trial demonstrated a 5-year OS of 26%, supporting the exclusion of surgery in appropriately selected patients. 12,13fortunately, despite encouraging survival outcomes, the rates of local-regional failure were >50%.While local-regional control may be   | 597 improved with esophagectomy, this does not necessarily correlate to improvement in OS. 17 Radiation therapy, even well planned and delivered, is associated with a variable response rate, which could optimistically be estimated as the pCR rate (15%-30%).The variability of pathologic surgical outcome for esophagectomy is relatively limited, with margin positive resection rates typically <10%, 2,3,8 suggesting that timely surgical resection is necessary to provide primary site control.
Our results identified both clinical N-and T-status as poor prognostic factors in our overall cohort, whereas T-status alone was predictive of mortality in the DE cohort.When patients in the DE cohort were grouped by N0 versus N > 0, no difference in OS was seen.These results would suggest that for patients with advanced primary tumors, radiation may be insufficient, thereby supporting esophagectomy to provide definitive local therapy.This result is supported by a prior review of the NCDB evaluating the outcomes of neoadjuvant CXRT for esophageal carcinoma, concluding that the benefit lay primarily with those individuals who were node-positive, while node-negative tumors failed to realize the same benefit. 18rther supporting surgery for definitive local control, a study by Spicer and colleagues demonstrated equivalent outcomes between neoadjuvant CXRT versus chemotherapy when a modified en-bloc esophagectomy technique was employed. 19ving a step further, in a multicenter study of DE, persistent as opposed to recurrent disease was associated with a significant reduction in both 3-year OS (40.9% vs. 56.2%)and disease-free state (36.6% vs. 51.6%).In context, omission of surgery may only be appropriate in those individuals with a reasonable expectation that radiation will adequately manage the primary tumor.Persistent disease after definitive CXRT should be viewed in a similar light to a positive surgical margin, that is, failure of the treatment modality.
Unfortunately, the ability to preoperatively predict pCR following CXRT is limited.Initial clinical staging is often no better, with clinical-pathologic staging discordance being substantial, often markedly for early/intermediate stage disease (cT2N0). 20In the absence of accurate clinical staging, appropriately selecting individuals for possible delayed primary surgical resection may be ill- Limitations to our study include its retrospective design and likely a significant selection bias related to the decision to pursue upfront resection as compared to a delayed primary resection or DE.
Our analysis in the timing of surgery after radiation therapy is limited as the NCDB does not collect data about the end date of chemotherapy.In addition, the database provides no information on the specific chemotherapy regimen, the number of cycles  | 599 Kaplan-Meier curve of delayed esophagectomy versus nondelayed esophagectomy for cN > 0 patients.T A B L E 4 Survival of delayed esophagectomy patients by clinical T stage and clinical N stage.

3 T A B L E 5
Timing of esophagectomy after completion of RT in overall cohort.
advised.Taken together, chemoradiation alone fails to provide consistent and complete local control for advanced primary tumors, even in the face of effective distant disease management with systemic chemotherapy.By failing to provide a complete therapeutic strategy, including esophagectomy, patients may be receiving an inadequate treatment.Additionally, DE may expose patients to an inappropriate risk of postoperative morbidity and mortality, which is converse to the rationale for application of a DE treatment paradigm.Finally, in our study, the multivariable Cox regression for the DE cohort identified treatment outside of an academic facility as predictive of mortality.As has been previously shown, esophagectomy outcomes are improved when performed at high-volume specialty centers.With the significant perioperative risk profile, the decision to place patients on a treatment pathway with the potential for DE should only occur in a setting where expertise in high-risk esophageal surgery is available.

F
I G U R E 3 Kaplan-Meier overall survival curve grouped by surgery timing.delivered, and the need for dose reduction or omission.Because of this paucity of information, no significant conclusions can be made about the efficacy of CXRT alone or the exact timing after completion of both chemotherapy and radiation therapy.Additionally, the NCDB data set in its current form does not provide granularity allowing for determination of the reason for DE, potentially leading to differential survival outcomes between treatment groups.Lastly, in the context of prior data supporting esophagectomy at high-volume centers, given that center type was noted to be significantly associated with mortality in the DE patients, these results may not be broadly applicable to all patients.5| CONCLUSIONOur study demonstrates that DE is associated with a decrease in OS as compared to NDE for patients with local-regionally advanced esophageal adenocarcinoma.Advancing T-status represents an independent risk factor for DE failure.Accurate clinical staging is imperative in deciding whether nonsurgical definitive therapy should be considered as opposed to trimodality treatment with extirpative surgery.These results question whether DE should be offered to medically fit patients with locally (T3/T4) advanced primary esophageal adenocarcinoma.
T A B L E 1