Prognosis after neoadjuvant chemoradiation or chemotherapy for locally advanced gastro-oesophageal junctional adenocarcinoma

Abstract

Background

Trials typically group cancers of the gastro-oesophageal junction (GOJ) with oesophageal or gastric cancer when studying neoadjuvant chemoradiation and perioperative chemotherapy, so the results may not be fully applicable to GOJ cancer. Because optimal neoadjuvant treatment for GOJ cancer remains controversial, outcomes with neoadjuvant chemoradiation versus chemotherapy for locally advanced GOJ adenocarcinoma were compared retrospectively.

Methods

Data were collected from all patients who underwent neoadjuvant treatment followed by surgery for adenocarcinoma located at the GOJ at a single high-volume institution between 2002 and 2017. Postoperative major complications and mortality were compared between groups using Fisher’s exact test. Overall survival (OS) and disease-free survival (DFS) were assessed by log rank test and multivariable Cox regression analyses. Cumulative incidence functions were used to estimate recurrence, and groups were compared using Gray’s test.

Results

Of 775 patients, 650 had neoadjuvant chemoradiation and 125 had chemotherapy. These groups were comparable in terms of clinical tumour and lymph node categories, although the chemoradiation group had greater proportions of white men, complete pathological response to chemotherapy, and smaller proportions of diffuse cancer, poor differentiation, and neurovascular invasion. Postoperative major complications (20.0 versus 17.6 per cent) and 30-day mortality (1.7 versus 1.6 per cent) were not significantly different between the chemoradiation and chemotherapy groups. After adjustment, type of therapy (chemoradiation versus chemotherapy) was not significantly associated with OS (hazard ratio (HR) 1.26, 95 per cent c.i. 0.96 to 1.67) or DFS (HR 1.27, 0.98 to 1.64). Type of recurrence (local, regional, or distant) did not differ after neoadjuvant chemoradiation versus chemotherapy.

Conclusion

In patients undergoing surgical resection for locally advanced adenocarcinoma of the GOJ, OS and DFS did not differ significantly between patients who had neoadjuvant chemoradiation compared with chemotherapy.

Cancer of the gastro-oesophageal junction (GOJ) is often grouped with either oesophageal or gastric cancer in trials, given its indistinct location in the distal oesophagus and proximal stomach. This retrospective study compared outcomes among 650 patients with adenocarcinoma of the GOJ treated with neoadjuvant chemoradiation and 125 who had neoadjuvant or perioperative chemotherapy. Neoadjuvant chemoradiation was not significantly associated with increased 30-day major complications or mortality, nor improved overall or disease-free survival, compared with perioperative chemotherapy.

Introduction

The diagnosis of cancer of the gastro-oesophageal junction (GOJ) comes with unique challenges given its location in the distal oesophagus and proximal stomach, and the lack of consensus on whether it should be treated as oesophageal or gastric cancer. The incidence rates of proximal gastric and oesophageal adenocarcinoma have risen over recent decades, and they are now the second and sixth leading causes of cancer-related deaths worldwide respectively¹. Surgery remains the cornerstone of curative therapy; neoadjuvant or perioperative treatment has been found to improve survival for patients with locally advanced oesophageal and gastric cancers. However, one of the largest challenges in managing GOJ cancer is that RCTs have demonstrated that neoadjuvant chemoradiation improves survival in oesophageal cancer, whereas perioperative chemotherapy improves survival in gastric cancer^2–4; no studies have yet reported exclusively on neoadjuvant treatment for adenocarcinoma of the GOJ. Even though trials in oesophageal or gastric cancer include tumours located at the GOJ, the number of patients is often too small for meaningful subgroup analysis. Among patients with GOJ adenocarcinoma receiving neoadjuvant treatment in the United States National Cancer Database, chemoradiation was the treatment of choice for 85 per cent⁵. Retrospective comparisons of patients treated with neoadjuvant chemoradiation and neoadjuvant chemotherapy have suggested that these treatments may be oncologically equivalent, despite higher rates of pathological complete response and negative surgical margins after chemoradiation which may result in improved local control^6–11.

Studying GOJ cancer is complicated by the difficulty in defining and classifying these lesions. Different trials have used selective definitions and even the widely used Siewert classification is prone to ambiguous interpretation¹². Yet the eighth edition of the AJCC Cancer Staging Manual for GOJ cancers is based on a modified Siewert classification: cancers with their epicentre within the proximal 2 cm of cardia (Siewert types I and II) are classified as oesophageal cancers, and those with epicentres more than 2 cm distal from the GOJ (Siewert type III) as gastric cancers¹³. As a result, neoadjuvant management of GOJ adenocarcinoma is often based on the cancer location as defined by the Siewert classification. Determining the Siewert classification is difficult. Before operation, the epicentre of the cancer is not always obvious to the clinician (gastroenterologist or surgeon). After surgery, the epicentre may elude clear determination by the pathologist because neoadjuvant treatment often obscures the proximal and distal borders of the carcinoma. Because the Siewert classification was described in an era when patients had surgery upfront, a precise nomenclature of Siewert I, II, or III types is not always agreed on after chemotherapy or chemoradiation; nevertheless, the designation of these tumours as GOJ cancers is approved collectively by most groups. As a result, the optimal management of adenocarcinomas of the GOJ should not be based on subtle and sometimes unclear borders above or below the Z-line. An alternative classification approach has been suggested by the recent molecular characterization of GOJ adenocarcinoma; direct comparison with both oesophageal adenocarcinoma and gastric adenocarcinoma showed GOJ tumours to be a combined entity of oesophageal-like (31 per cent) and gastric-like (69 per cent) cancers, without a unique molecular cluster¹⁴. For all these reasons, GOJ adenocarcinomas were evaluated as a common disease entity.

The aim of this study was to investigate, among patients with resected locally advanced GOJ adenocarcinoma, whether outcomes differed between those treated with neoadjuvant chemoradiation and those who received neoadjuvant or perioperative chemotherapy.

Methods

A single-institution, retrospective cohort study was undertaken after approval from the Memorial Sloan Kettering Cancer Centre (MSK) institutional review board. All patients with non-metastatic disease who underwent resection for a primary adenocarcinoma located at the GOJ, cardia, or upper stomach, as concluded by the pathologist on the surgical specimen, between January 2002 and December 2017, were selected for review from two prospectively maintained databases containing all gastric and oesophageal cancer resections at MSK. If the cancer was classified as a tumour of the GOJ (Siewert I, II, or III) by an experienced pathologist, the patient was eligible for analysis (Fig. 1). If the cancer location was classified as cardia or upper stomach, the pathologist’s detailed description of the surgical specimen was reviewed. If the epicentre or upper border of the tumour was within 2 cm below the GOJ, the patient was deemed eligible. To prevent misclassification of large oesophageal or gastric cancers that infiltrated the GOJ, Siewert I cancers extending beyond 5 cm above the GOJ and Siewert III cancers extending beyond 5 cm below the GOJ were excluded. Additionally, patients who did not have neoadjuvant or perioperative treatment, and those who received chemoradiation with neoadjuvant intent and achieved a clinical complete response, so did not proceed to surgery, were excluded. For simplification, perioperative treatment is referred to as neoadjuvant treatment throughout this article, meaning that some patients who received neoadjuvant treatment also had additional adjuvant treatment.

Fig. 1.

Study flow chart

GOJ, gastro-oesophageal junction

Data on clinicopathological and treatment characteristics and follow-up were collected from the databases and electronic medical records. Clinical and pathological tumour and lymph node status were defined according to the AJCC eighth edition of the TNM staging system guidelines for the oesophagus and oesophagogastric junction¹⁵. Information on differentiation grade, vascular invasion, and neural invasion was retrieved from the postoperative pathological evaluation and from the biopsy notes if there was no residual cancer. Pathological evaluation after gastrectomy and oesophagectomy was performed by the same expert gastrointestinal pathologists. The Siewert classification was defined according to the National Comprehensive Cancer Network (NCCN) guideline for oesophageal and oesophagogastric junction cancers and gastric cancer^16,17 on the postoperative pathological GOJ specimen. If there was no residual tumour, the centre of the scar was used as proxy for the tumour centre. Postoperative complications were defined by the modified Clavien–Dindo classification; major complications were defined as those of grade III–V¹⁸.

Treatment, follow-up, and recurrence

In general, patients with cT3 or cT4 tumours or clinically positive lymph nodes are considered for neoadjuvant treatment at MSK. If patients received chemoradiation, they were offered induction chemotherapy followed by chemoradiation, with a duration of approximately 10–12 weeks, followed by surgery after 5–8 weeks. The chemotherapy regimens included cisplatin or irinotecan up to 2010¹⁹, then paclitaxel and carboplatin based on the CROSS (The Dutch ChemoRadiotherapy for Oesophageal cancer followed by Surgery Study) trial²⁰, and FOLFOX (leucovorin, 5-fluorouracil, oxaliplatin) from 2017 onwards based on the CALGB 80803 trial (A randomized phase II trial of PET scan-directed combined modality therapy for esophageal cancer)²¹. The response has been assessed after induction chemotherapy by PET since 2010. The standard radiation dose was 50.4 Gy in 28 fractions. If patients received neoadjuvant or perioperative chemotherapy, the duration was approximately 6 weeks followed by surgery after 4–6 weeks. The chemotherapy regimens contained epirubicin based on the MAGIC trial (Medical Research Council Adjuvant Gastric Infusional Chemotherapy)³, or FLOT (fluorouracil plus leucovorin, oxaliplatin, and docetaxel) based on the FLOT4 trial (Perioperative chemotherapy with docetaxel, oxaliplatin, and fluorouracil/leucovorin (FLOT) versus epirubicin, cisplatin, and fluorouracil or capecitabine (ECF/ECX) for resectable gastric or gastroesophageal junction (GEJ) adenocarcinoma)²². Patients with a clinical complete response were not offered surveillance, and surgery was strongly recommended.

After treatments, including surgery, patients were followed up in the outpatient setting by the surgeon or medical oncologist every 3–6 months in the first 2 years and every 6–12 months thereafter. Outpatient visits included physical examination, laboratory tests, and radiographic imaging, most commonly CT of the chest, abdomen, and pelvis. Recurrence was classified as local, regional, or distant, and patients were categorized based on the first recurrence detected at the furthest anatomical site. For example, recurrence in a patient with both local and regional recurrence was classified as regional, whereas recurrence in a patient with both regional and distant recurrence was classified as distant. Date of recurrence was defined as the date of pathological confirmation or, if no pathological confirmation was obtained, the date of radiological imaging that was (highly) suspicious for recurrence and led to changes in management.

Statistical analysis

Clinicopathological characteristics and postoperative complications were compared between treatment strategy groups (neoadjuvant chemotherapy versus chemoradiation) by χ² or Fisher’s exact test for categorical variables, and Wilcoxon rank-sum test for continuous variables. Overall survival (OS) was calculated from the date of surgery until date of death or last follow-up for survivors. Disease-free survival (DFS) was defined as survival until disease recurrence at any site or death from any cause. Univariable and multivariable survival analysis was performed by Cox regression analysis; the multivariable analysis included relevant clinical variables that are considered possible confounders. Survival curves were estimated by the Kaplan–Meier method and compared using the log rank test; median follow-up was calculated using the reverse Kaplan–Meier approach. Cumulative incidence functions were used to estimate local, regional, and distant recurrences, where death without recurrence was considered a competing risk. Groups were compared using Gray’s test. All statistical tests were two-sided; P < 0.050 was considered to indicate statistical significance. Statistical analyses were undertaken using R version 3.6 (R Foundation for Statistical Computing, Vienna, Austria).

Results

A total of 775 patients who underwent resection of a GOJ adenocarcinoma were included (Fig. 1); 650 (83.9 per cent) were treated with neoadjuvant chemoradiation and 125 with neoadjuvant chemotherapy. Compared with the chemotherapy group, patients who had chemoradiation were more likely to be men (85.1 versus 72.0 per cent; P < 0.001) and white (93.7 versus 87.2 per cent; P = 0.018), whereas the two groups were comparable in age (median 63 versus 62 years), clinical T category (cT3–4 in 90 per cent in both groups), and clinical lymph node status (cN-positive in 67.1 versus 57.6 per cent) (Table 1). Oesophagectomy was performed in 633 patients (97.4 per cent) treated with chemoradiation and 45 (36.0 per cent) who had chemotherapy, whereas 17 (2.6 per cent) and 80 (64.0 per cent) respectively underwent gastrectomy (Table 2). Patients who had chemoradiation were more likely to have a low pathological tumour category after treatment (ypT0: 20.3 versus 8.0 per cent; P < 0.001) and negative lymph node status (ypN0: 56.9 versus 42.4 per cent; P < 0.001), less likely to have perineural (28.5 versus 50.4 per cent; P < 0.001) and vascular (28.5 versus 56.0 per cent; P < 0.001) invasion, but also less likely to have diffuse cancer type (6.8 versus 13.6 per cent; P < 0.001) and poor differentiation grade (36.8 versus 52.8 per cent; P = 0.002). The median number of lymph nodes dissected was lower in the chemoradiation group (21 versus 23; P = 0.003), and the proportion of patients with negative margins was not significantly different from that in the chemotherapy group (96.0 versus 92.8 per cent; P = 0.179) (Table 2).

Table 1.

Patient demographics and clinical characteristics

	Chemotherapy only (n = 125)	Chemoradiation (n = 650)	P †
Age (years)*	62 (55–68)	63 (57–70)	0.370‡
Sex ratio	90 : 35	553 : 97	<0.001
Race			0.018
White	109 (87)	609 (94)
Other	16 (13)	41 (6.3)
cT category			>0.900
cT1–T2	12 (10)	61 (9.6)
cT3–T4	105 (90)	577 (90)
Unknown	8	12
cN category			0.115
cN0	49 (40)	211 (33)
cN+	72 (60)	436 (67)
Unknown	4	3

Values in parentheses are percentages unless indicated otherwise;

*values are median (i.q.r.).

^†χ² test, except

^‡Wilcoxon rank-sum test.

Table 2.

Pathological and treatment characteristics

	Chemotherapy only (n = 125)	Chemoradiation (n = 650)	P †
Surgery type			<0.001
Oesophagectomy	45 (36)	633 (97)
Gastrectomy	80 (64)	17 (2.6)
ypT category			<0.001
ypT0	10 (8.0)	132 (20)
ypT1	8 (6.4)	125 (19)
ypT 2	21 (17)	118 (18)
ypT3	77 (62)	264 (41)
ypT 4	9 (7.2)	11 (1.7)
ypN category			<0.001
ypN0	53 (42)	370 (57)
ypN1	30 (24)	157 (24)
ypN2	24 (19)	94 (14)
ypN3	18 (14)	29 (4.5)
Siewert classification			<0.001
1	9 (7.2)	129 (20)
2	51 (41)	432 (67)
3	65 (52)	89 (14)
Laurén classification			<0.001
Intestinal or missing	89 (71)	552 (85)
Diffuse	17 (14)	44 (6.8)
Mixed	19 (15)	54 (8.3)
Differentiation grade			0.002
Poor	66 (53)	239 (37)
Moderate	44 (35)	279 (44)
Well	5 (4.0)	14 (2.2)
No residual tumour	10 (8.0)	107 (17)
Unknown	0	11
Neural invasion	63 (52)	185 (29)	<0.001
Unknown	5	5
Vascular invasion	70 (58)	185 (29)	<0.001
Unknown	4	3
Total no. of positive lymph nodes*	1 (0–4)	0 (0–2)	<0.001‡
Unknown	0	8
Total no. of lymph nodes harvested*	23 (18–32)	21 (16–27)	0.003‡
Unknown	0	8
Surgical margin			0.179
Negative	116 (93)	624 (96)
Positive	9 (7.2)	26 (4.0)

Values in parentheses are percentages unless indicated otherwise;

*values are median (i.q.r.).

^†χ² or Fisher’s exact test, except

^‡Wilcoxon rank-sum test.

Postoperative results and follow-up

The chemoradiation and chemotherapy groups had similar rates of postoperative major complications (20.0 versus 17.6 per cent; P = 0.542), including number of deaths within 30 days of surgery (11 (1.7 per cent) versus 2 (1.6 per cent); P > 0.900). After a median follow-up of 5.5 (95 per cent c.i. 5.2 to 6.0) years, 361 patients had developed disease recurrence, of whom 305 had died; an additional 151 patients had died without a known recurrence.

The 5-year OS rate in the total cohort was 41 (95 per cent c.i. 38 to 45) per cent, 40 (36 to 45) per cent among patients who had chemoradiation, and 45 (36 to 56) per cent among those treated with chemotherapy (P = 0.146) (Fig. 2). After adjustment for age, sex, race, cT category, cN category, and surgical margin status in the multivariable analysis, OS for patients who had chemoradiation was comparable to that for patients who received chemotherapy (hazard ratio (HR) 1.26, 95 per cent c.i. 0.95 to 1.65; P = 0.106) (Table 3).

Fig. 2.

Overall survival in patients who had neoadjuvant chemoradiation or chemotherapy followed by resection for adenocarcinoma of the gastro-oesophageal junction

P = 0.15 (log rank test).

Table 3.

Multivariable Cox regression analysis of overall and disease-free survival

	Overall survival		Disease-free survival
	Hazard ratio	P	Hazard ratio	P
Neoadjuvant treatment
Chemotherapy only	1.00 (reference)		1.00 (reference)
Chemoradiotherapy	1.26 (0.95, 1.65)	0.106	1.26 (0.97, 1.64)	0.081
Age (per year)	1.02 (1.01, 1.03)	<0.001	1.01 (1.01, 1.02)	0.001
Sex
M	1.00 (reference)		1.00 (reference)
F	1.02 (0.80, 1.31	0.869	1.02 (0.81, 1.29)	0.856
Race
White	1.00 (reference)		1.00 (reference)
Other	0.99 (0.67, 1.46	>0.900	0.94 (0.66, 1.35)	0.744
cT category
cT1–T2	1.00 (reference)		1.00 (reference)
cT3–T4	1.01 (0.73, 1.40	>0.900	0.98 (0.72, 1.34)	>0.900
cN category
cN+	1.00 (reference)		1.00 (reference)
cN–	0.82 (0.67, 1.00)	0.052	0.79 (0.65, 0.95)	0.014
Year of surgery	0.98 (0.96, 1.00)	0.110	0.99 (0.97, 1.01)	0.302
Surgical margin
Negative	1.00 (reference)		1.00 (reference)
Positive	2.91 (1.94, 4.36)	<0.001	2.41 (1.62, 3.57)	<0.001

Values in parentheses are 95 per cent confidence intervals.

The cumulative incidence of disease recurrence did not differ significantly between the two groups (P = 0.859); similar relationships were observed at local (P = 0.290), regional (P = 0.150), and distant (P = 0.178) sites (Fig. 3). The 5-year cumulative incidence of local recurrence was 5 (95 per cent c.i. 4 to 7) versus 7 (3 to 13) per cent in the chemoradiation versus chemotherapy groups; 5-years regional recurrence rates were 7 (5 to 9) versus 11 (6 to 17) per cent, and 5-year distant recurrence rates were 36 (32 to 40) versus 30 (22 to 39) per cent. Similar to OS, DFS did not differ significantly between the chemoradiation and chemotherapy groups (HR 1.22, 0.95 to 1.56; P = 0.119), nor after adjustment in the multivariable analysis (HR 1.26, 0.97 to 1.64; P = 0.081) (Table 3). Five-year DFS rates were 34 and 39 per cent respectively (Fig. 4).

Fig. 3.

Cumulative incidence of disease recurrence in patients who had neoadjuvant chemoradiation or chemotherapy followed by resection for adenocarcinoma of the gastro-oesophageal junction

a Local, b regional, and c distant recurrence. a P = 0.29, b P = 0.15, c P = 0.178 (Gray’s test).

Fig. 4.

Disease-free survival in patients who had neoadjuvant chemoradiation or chemotherapy followed by resection for adenocarcinoma of the gastro-oesophageal junction

P = 0.12 (log rank test).

Discussion

In this cohort of patients with locally advanced GOJ adenocarcinoma who underwent surgical resection with curative intent, OS and DFS did not differ significantly between patients who had neoadjuvant chemoradiation and those who received neoadjuvant/perioperative chemotherapy.

Previous studies yielded limited evidence on the optimal approach for GOJ cancers because most focused either on oesophageal cancer or on gastric cancer. For oesophageal cancer, the rationale for neoadjuvant chemoradiation over surgery alone (as recommended by the latest NCCN guidelines) is based mainly on the results of the CROSS trial²⁰, which enrolled patients with adenocarcinoma or squamous cell carcinoma of the oesophagus. That trial showed improved survival for neoadjuvant chemoradiation plus surgery compared with surgery alone, especially in patients with squamous cell carcinoma. Trials of neoadjuvant chemotherapy in oesophageal cancer have shown mixed results^23,24. For gastric cancer, the rationale for perioperative chemotherapy over surgery alone originated mainly from the MAGIC trial³, which showed improved survival for perioperative chemotherapy plus surgery over surgery alone in patients with gastric adenocarcinoma. However, the CROSS trial included only 88 (of 368) patients with GOJ cancer, and the MAGIC trial only 58 (of 503). A French trial⁴ of perioperative chemotherapy included more GOJ cancers (144 of CT 224) and found a survival benefit for chemotherapy over surgery alone. Further survival benefit has been shown in trials using other chemotherapy regimens. The FLOT4 trial²², which included 398 (of 716) patients with GOJ cancer, examined perioperative FLOT versus epirubicin and cisplatin plus fluorouracil (ECF) or plus capecitabine (ECX) for gastric or GOJ adenocarcinoma. The investigators found that FLOT improved survival over the regimen used in the MAGIC trial (ECF).

To date, only one phase III trial (POET trial)²⁵ has compared neoadjuvant chemoradiation and chemotherapy in GOJ adenocarcinoma exclusively. This trial was unfortunately closed prematurely owing to insufficient accrual, but the interim analysis did not show a survival advantage for chemoradiation²⁵. Two smaller trials^26,27 directly compared neoadjuvant chemoradiation and chemotherapy in oesophageal cancer, including GOJ cancers, but found no difference in survival. The NeoRes trial²⁶ included 31 patients with GOJ cancer of a total of 181. Currently, there are four ongoing trials comparing neoadjuvant chemoradiation and chemotherapy followed by surgery in either oesophageal or gastric cancer, and patients with GOJ tumours are eligible for inclusion in all four. The CRITICS-II trial²⁸ is randomizing patients with gastric cancer into three different preoperative treatment groups: chemotherapy, chemotherapy followed by chemoradiation, and chemoradiation. Neo-AEGIS (NEOadjuvant trial in Adenocarcinoma of the oEsophagus and oesophagoGastric junction International Study)²⁹ is randomizing patients with oesophageal adenocarcinoma to a modified MAGIC regimen or the CROSS regimen. The TOPGEAR trial³⁰ is randomizing patients with gastric adenocarcinoma to perioperative chemotherapy similar to that in the MAGIC trial or to this regimen plus preoperative chemoradiation. Finally, the ESOPEC trial³¹ is randomizing patients with oesophageal adenocarcinoma to the FLOT or CROSS regimen. The results of these trials are eagerly anticipated.

In previous trials, the addition of radiotherapy to chemotherapy resulted in more downstaging^25,26,27. Indeed, in the present study, pT and pN categories were lower after treatment in the chemoradiation group. However, this study did not demonstrate a statistically significant difference between the chemotherapy and chemoradiation groups in terms of negative surgical margins, although other retrospective studies have^25,27,32. The significantly better pT and pN categories in the chemoradiation group were not accompanied by better survival, suggesting that downstaging is not associated with improved survival in GOJ adenocarcinoma. The same was observed in the NeoRes trial, and in five multicentre, retrospective analyses (2 studies from the US National Cancer Database, 1 from the Swedish National Cancer Database, 1 combining 3 large-volume US centres, and 1 study combining 10 European centres) in patients with oesophageal and GOJ cancer treated with neoadjuvant chemoradiation and chemotherapy^5,9,10,25. Despite significant post-treatment pathological downstaging of tumour and lymph nodes in the chemoradiation group, no difference was seen in 3-year OS or DFS^9,10. Similarly, a meta-analysis⁶ of mostly retrospective studies comparing neoadjuvant chemoradiation and chemotherapy in GOJ adenocarcinoma found a higher pathological complete response rate in the chemoradiation group, but no unadjusted OS advantage. Although a more frequent pathological complete response is often used as an argument for preferring chemoradiation over chemotherapy alone, the present findings, as well as results of previous studies, suggest that more downstaging is not associated with improved survival in GOJ adenocarcinoma.

In the meta-analysis⁶ comparing neoadjuvant chemoradiation and chemotherapy in GOJ adenocarcinoma, chemoradiation was associated with a reduced risk of locoregional relapse. In contrast, no significant difference in the cumulative incidence of local or regional recurrences was found in the present study, or in an investigation from three large-volume US centres reported by Spicer and colleagues¹⁰. The factors independently associated with OS and DFS in the present study were age, cN status, and surgical margin status. Associations between positive surgical margins and worse survival are often used as an argument for the addition of radiotherapy. However, the positive margin rate in the chemotherapy group was not clinically or statistically significantly different from that in the chemoradiation group (7 versus 4 per cent).

Although there has been much concern that chemoradiation might increase postoperative complication rates, this was not observed in the present study. The major complication rate (20 versus 17 per cent) and 30-day mortality rate (1.7 versus 1.6 per cent) was similar in the chemoradiation and chemotherapy groups. Similar complication and 30-day mortality rates were observed in the study of three US large-volume centres¹⁰. In contrast, a large European study⁷ reported a higher anastomotic leak rate (23 versus 6.8 per cent) and higher absolute mortality rate (4.1 versus 1.4 per cent) in the chemoradiation group. The European study⁷ also noted a large difference in number of lymph nodes harvested in the chemoradiation versus chemotherapy group (14 versus 27), and lymph node harvest was associated with DFS in the chemotherapy group only. In the present cohort, the difference in median number of lymph nodes resected was minimal (21 for chemoradiation group versus 23 for chemotherapy group), but statistically significant (P = 0.003). The low positive margin rate in the chemotherapy group and low major complication rate in the chemoradiation group may reflect the multimodal care at a tertiary-care institution. These results may not apply to other settings. On the other hand, long-term adverse events were not reported here. The NeoRes trial²⁶ reported an increase in mortality rate from 1 versus 0 per cent at 30 days to 8 versus 3 per cent at 90 days in the chemoradiation versus chemotherapy group. Although the mortality rate was not significantly different between the chemoradiation and chemotherapy groups, the trial results suggest that morbidity and mortality rates might differ beyond 30 days, but these data were not collected.

The retrospective nature of this analysis comes with disadvantages, such as the risk of selection bias based on disease stage at presentation. However, age and clinical TNM stage were similar in the chemoradiation and chemotherapy groups. To attempt to control for confounding, a multivariable survival analysis was performed, adjusting for variables including age and clinical TNM stage. However, residual unmeasured confounding factors cannot be ruled out. The differences between the groups in sex and race were expected, as it is known that white men are over-represented among patients with oesophageal cancer, but somewhat less so among those with gastric cancer. GOJ cancers located higher in the oesophagus were more likely to be treated with chemoradiation, according to current treatment standards, and so the chemoradiation group included more white men than the chemotherapy group. Similarly, the equivalent outcomes of the two treatments may reflect this influence of GOJ cancer position on choice of treatment, rather than reflecting true equivalence of the treatments. At MSK, GOJ cancers are treated by both thoracic surgeons performing oesophagectomy and gastric cancer surgeons undertaking gastrectomy. However, differences in surgical technique were not the focus of the present study, and it was assumed that they do not influence prognosis if performed according to current standards³³. Another disadvantage is the lack of a larger denominator; the number of patients who started neoadjuvant treatment but did not go on to surgery with curative intent is unknown. It can be assumed that the proportion of patients who did not proceed to surgery with curative intent because of disease progression, patient’s wish, or treatment toxicity was similar in the two treatment groups, as the drop-off rate was 5.8 per cent in the CROSS trial and 3.1 per cent in the FLOT trial. As surveillance is not offered to patients with a clinical complete response at this institution, the only reasons for stopping treatment before surgery would have been disease progression, patient preference, or treatment toxicity. Furthermore, comparisons of present findings with others may be limited by differences in definition of GOJ cancers, and by differing interpretations of the Siewert classification in the literature and in clinical practice. In contrast to all previously published retrospective studies, a strength of the present study is the review of detailed cancer location descriptions of the postoperative pathological specimen to ensure the inclusion of GOJ cancers according to the definition of the eighth edition of the AJCC Cancer Staging Manual. Having studied GOJ adenocarcinoma as a single entity, without stratifying for Siewert types I–III, is considered a strength because it emphasizes the goal of identifying the optimal treatment regardless of an anatomical classification that is used ambiguously.