Long-term outcome after perioperative chemotherapy and surgery for gastro-esophageal adenocarcinoma

Abstract

Background

The long-term fate of patients treated for resectable gastro-esophageal adenocarcinoma with perioperative chemotherapy outside randomized clinical trials (RCTs) is poorly described. In this national cohort, we report on outcomes after 12 years.

Materials and methods

Baseline clinicopathological factors and blood tests were collected in 285 patients treated from May 2008 to June 2010, and postsurgical factors were collected in the 202 patients that were radically resected. Response to preoperative chemotherapy was assessed by postsurgical restaging. Additional information on second cancers, comorbidities, and competing causes of death was obtained.

Results

Overall survival (OS) at 5 and 10 years was 31.9% and 24.2%, respectively. Multivariate analysis (MA) showed prognostic value of clinical T- and N-stage, dysphagia, and Charlson Comorbidity Index. Elevated leucocytes and lactate dehydrogenase, and low lymphocytes were additional adverse prognostic factors. Ten-year incidence rate of second cancers was 10.1%.

OS at 5 and 10 years from radical surgery was 43.1% and 32.1%, respectively. MA showed prognostic value of postneoadjuvant pathological (yp) Union for International Cancer Control (UICC) stage and downstaging. However, downstaging was strongly correlated with clinical stage. At the 5-year landmark, marginally significant trends toward a favorable recurrence-free survival were associated with ypT0 and age <70 years, but only 3.4% experienced late recurrences. Postoperative complications caused a quarter of deaths in the half of radically resected patients who died without recurrence.

Conclusions

Five-year OS was 4%-6% points inferior to RCTs. Pretreatment white blood cell counts, but not postoperative, could supplement clinical prognostic factors. Downstaging by preoperative chemotherapy was prognostic but correlated strongly with pretreatment clinical stage.

Highlights

• •National, 12-year outcomes of perioperative chemotherapy and resection for gastro-esophageal adenocarcinoma were reported.
• •Survival at 5 and 10 years was 32% and 24%. Five-year survival was 4%-6% points inferior to randomized clinical trials.
• •Post-operative complications caused a quarter of deaths in radically resected patients who died without recurrence.
• •Pretreatment white blood cell counts, but not post-operative, could supplement clinical prognostic factors.
• •Downstaging by chemotherapy was prognostic but strongly correlated with pretreatment stage.

Introduction

Gastro-esophageal adenocarcinoma is a global health problem with a high mortality rate.¹ Curative intent treatment, usually surgery, can be offered to ∼40% of patients,² but is associated with a high risk of recurrence and death.³

Fifteen years ago, randomized studies demonstrated an improved survival by adding perioperative chemotherapy to surgery in patients with resectable disease. In the pivotal MAGIC study, perioperative treatment with epirubicin, cisplatin and fluorouracil (ECF) resulted in an increased curative resection rate, and the 5-year survival rate rose significantly from 23% to 36%.⁴ Similarly, in the ACCORD-07 trial, treatment with cisplatin and fluorouracil resulted in an improvement of 5-year survival from 24% to 38%.⁵ The FLOT4 trial, published in 2019, showed a further improvement in 5-year survival from 36% to 45% with perioperative docetaxel, oxaliplatin and fluorouracil (FLOT) compared with ECF or epirubicin, oxaliplatin and capecitabine (EOX).⁶

Preoperative chemoradiation with adjuvant checkpoint inhibition in patients with incomplete pathological response was another equally recommended option in patients with cancers located in the esophagus or at the gastro-esophageal junction (GEJ).^7,8 However, recently published results of the ESOPEC trial showed that FLOT leads to improved overall survival (OS) compared with chemoradiation,⁹ and for the vast majority of patients with gastro-esophageal adenocarcinomas, perioperative chemotherapy with FLOT is now considered standard of care.¹⁰

While the pivotal studies changed clinical practice, patients included in clinical trials are often highly selected.¹¹ Reassuringly, population-based studies showed an increased survival among curatively treated patients after implementation of perioperative chemotherapy,¹² and early toxicity, adherence to treatment, predictive factors for not reaching resection and short-term prognostic factors have been described.^13,14 Still, relatively little is known of the long-term fate of off-trial patients.

In the current analysis of a complete national cohort of patients with resectable gastro-esophageal adenocarcinoma, the almost complete follow-up with a minimum 12-year-long observation time and uniform oncological treatment were key elements that are not usually exposed in studies. We wanted to assess prognostic factors for long-term survival, both before treatment start and after radical surgery. Finally, we wanted to investigate the occurrence of second primary cancers, as well as comorbidities and competing causes of death in patients without relapse.

Material and methods

Patients and data

The cohort included all patients diagnosed with resectable gastro-esophageal adenocarcinoma from 1 May 2008 to 29 June 2010, who initiated treatment with perioperative chemotherapy at all four treating hospital units in Denmark. Patients were identified from the Danish Esophago-Gastric Database (DEGD)¹⁵ supplemented with records at the oncological departments. Clinical, biochemical and pathological data were retrieved from electronic health records (EHRs), and supplementary pathological and biochemical data were retrieved from national and regional registers, respectively. We followed the European Society of Medical Oncology Guidance for Reporting Oncology Real-World Evidence (ESMO-GROW).¹⁶

Information on preoperative chemotherapy included treatment regimen, start and stop dates of chemotherapy, and reductions in dose density (treatment delay, abortion or dose reductions). Data on synchronous and metachronous second cancers, diagnosed from 2 months before start of chemotherapy and onward, were registered from the national pathology register. Performance status (PS) was scored according to the Eastern Cooperative Oncology Group (ECOG). Charlson Comorbidity Index (CCI) was calculated from age and comorbidities registered in EHRs,¹⁷ excluding cancer diagnoses. Staging of tumor (T) and regional nodal (N) sites, and overall stage was carried out according to Union for International Cancer Control (UICC) version 6.¹⁸ We also registered whether an [¹⁸F]2-fluoro-2-deoxy-D-glucose (FDG) positron emission tomography (PET) scan was used for clinical staging, whether endoluminal stenting had been carried out at baseline, and whether dysphagia was registered at treatment start. Finally, we collected information on routine blood tests before treatment start and after surgery.

All patients were planned for intended curative surgery with minimum D2 resection, according to highest applicable standard after national centralization of surgery to four centers.^19,20 The cohort of patients who received radical surgery was subject to further analysis of factors predicting relapse and survival from the date of resection. Patients with recurrence within this cohort were identified from EHRs. We collected information on number of series of chemotherapy administered postoperatively, and of causes of no postoperative chemotherapy. The effect of preoperative chemotherapy was assessed by the numerical difference between clinical and pathological UICC stage, and downstaging was categorized into complete [i.e. pathological complete response (pCR)], incomplete or no downstaging including progression.^21,22 We calculated the weight change from start on preoperative chemotherapy to identify patients with postsurgical weight loss >10%. Furthermore, we collected information from EHRs on added serious comorbidities since surgery. For those who died without recurrence, we obtained information on cause of death from EHRs (88%) or death certificates (12%).

Statistics

The date of chemotherapy initiation and date of resection were index dates for the main cohort and the cohort of radically resected patients, respectively. No patients were lost to follow-up. Cox regression analysis was used to assess factors associated with outcome. Medians or normal limits were used for analysis of blood tests. Firstly, univariable analyses were conducted with outcome as the dependent variables and each of the baseline characteristics as the independent variable. Secondly, multivariable logistic regression analysis was conducted and included factors significantly associated with outcome in the univariable analysis. A 5-year landmark analysis was carried out to assess factors predicting residual outcome. Wald test P values and profile likelihood confidence limits were reported. Schoenfeld residuals revealed no significant nonproportionality in the multivariable model, indicating that the assumption of proportional hazards was reasonable. Median survival and point estimates with 95% confidence intervals (CIs) were reported.

Ethical considerations

There was no requirement by Danish law for patient consent or notification of the study to the Scientific Ethics Committee. Registration number at Region North Denmark was 2021-037211. Data were collected from each region through cooperation agreements and entered from each region into a secure, unified REDCap® database that was created for the purpose. The General Data Protection Regulation and the Data Protection Act were complied with.

Results

Patients and treatment

We identified 293 patients fulfilling the inclusion criteria. In one patient diagnosis was changed to neuroendocrine carcinoma, and for 7 patients data were missing, leaving 285 patients for further analysis. Six patients included had mixed adenosquamous cell-type histology, and the four patients in stage IV had nonmetastatic gastric cancer. The median follow-up after initiation of preoperative chemotherapy was 2.2 years, and median follow-up after surgery was 2.8 years. Patients alive were followed up for at least 12 years from the date of initiation of preoperative chemotherapy.

Patients were initiated on perioperative chemotherapy with cisplatin (N = 151) or oxaliplatin (N = 133) together with epirubicin and flourouracil/capecitabine, while one patient received oxaliplatin and fluorouracil (FLOX). Blood test results were available for 161 (56.6%) patients at baseline. Except for a higher fraction of patients <60 years in the cohort with blood tests available (P = 0.02), baseline variables were evenly distributed among patients with available or unavailable blood tests (P ≥ 0.14) (Supplementary Table S1, available at https://doi.org/10.1016/j.esmogo.2025.100183), and the OS outcomes were similar (P = 0.12).

Prognostic factors at preoperative chemotherapy

The median OS from start of chemotherapy was 2.2 years [interquartile range (IQR) 0.94-9.7 years], and OS at 2, 5 and 10 years was 53.0%, 31.9% and 24.2%, respectively. Baseline variables and results of survival analysis are shown in Table 1. In univariate analysis, age >70 years, impaired ECOG PS, increasing CCI, endoluminal stenting, dysphagia, adenosquamous histology, and high T-, N-, and UICC stages were significantly associated with poor prognosis (P < 0.05), while the outcome of patients staged by FDG-PET showed a trend toward improvement (P = 0.05). In multivariate analysis CCI, dysphagia, T- and N-staging remained significant (P ≤ 0.04).

Table 1.

Baseline variables and results of overall survival analysis of 285 patients initiating perioperative chemotherapy for resectable gastro-esophageal adenocarcinoma

			Overall survival
			Univariate analysis		Multivariate analysisa
Variableb	Cut point	Patients (n)	HR (95% CI)	P value	HR (95% CI)	P value
Sex	Female	51	Ref.	—	—	—
	Male	234	1.20 (0.85-1.69)	0.31	—	—
Age, years	<60	103	Ref.	—	Ref.	—
	60-69	111	1.00 (0.74-1.36)	1.00	0.85 (0.61-1.17)	0.31
	70+	71	1.50 (1.08-2.08)	0.02	1.27 (0.89-1.82)	0.18
ECOG PS (8)	0	189	Ref.	—	Ref.	-
	1/2	86/2	1.43 (1.08-1.89)	0.01	1.11 (0.82-1.51)	0.50
CCI	0	174	Ref.	—	Ref.	-
	1	67	1.45 (1.06-1.97)	0.02	1.37 (0.99-1.90)	0.05
	2/3+	44	1.75 (1.23-2.51)	0.002	2.05 (1.37-3.07)	<0.001
BMI (11)	Normal	100	Ref.	—	—	—
	High	165	0.94 (0.71-1.24)	0.66	—	—
	Low	9	1.34 (0.65-2.77)	0.43	—	—
Endoluminal stenting	No	237	Ref.	—	Ref.	—
	Yes	48	1.64 (1.17-2.29)	0.004	1.26 (0.88-1.82)	0.21
Dysphagia (1)	Not present	78	Ref.	—	Ref.	—
	Present	206	1.62 (1.20-2.20)	0.002	1.61 (1.14-2.26)	0.006
Tumor site	Esophagus/GEJ	41/194	Ref.	—	—	—
	Stomach	50	0.87 (0.62-1.23)	0.44	—	—
Tumor type	Adenocarcinoma	279	Ref.	—	Ref.	—
	Adenosquamous carcinoma	6	2.29 (1.01-5.16)	0.046	1.10 (0.43-2.83)	0.84
T-stage (5)	1/2	2/67	Ref.	—	Ref.	—
	3	193	1.56 (1.13-2.15)	0.006	1.44 (1.03-2.02)	0.03
	4	18	1.48 (0.83-2.63)	0.19	1.39 (0.75-2.57)	0.29
N-stage (4)	0	70	Ref.	—	Ref.	—
	1	204	1.51 (1.10-2.08)	0.01	1.42 (1.01-1.99)	0.04
	2	7	2.35 (1.06-5.22)	0.04	2.80 (1.21-6.49)	0.02
UICC stage (5)	I	12	Ref.	—	—	—
	II	98	1.61 (0.74-3.50)	0.23	—	—
	III/IV	166/4	2.35 (1.10-5.01)	0.03	—	—
Staging with FDG-PET	Yes	73	Ref.	—	—	—
	No	212	1.35 (0.99-1.82)	0.05	—	—
Preoperative CTx regimen	EOFc	134	Ref.	-	—	—
	ECF	151	1.12 (0.87-1.46)	0.38	—	—
Subcohort of 161 patients with available blood testsd
Hemoglobin	Normal	106	Ref.	—	—	—
	<LLN	55	1.12 (0.79-1.61)	0.52	—	—
Total leucocytes [7.9 × 109/l]	≤Median	85	Ref.	—	Ref.	—
	>Median	76	1.53 (1.08-2.15)	0.02	1.86 (1.22-2.82)	0.004
Lymphocytes 1.9 × 109/l]	≤Median	82	Ref.	—	Ref.	—
	>Median	79	0.65 (0.46-0.92)	0.01	0.48 (0.29-0.81)	0.006
Neutrophils [4.87 × 109/l]	≤Median	81	Ref.	—	—	—
	>Median	80	1.25 (0.88-1.75)	0.21	—	—
NLR [2.56]	≤Median	81	Ref.	—	Ref.	—
	>Median	80	1.63 (1.16-2.30)	0.005	0.95 (0.58-1.56)	0.85
Monocytes (1) [0.605 × 109/l]	≤Median	80	Ref.	—	—	—
	>Median	80	1.16 (0.83-1.64)	0.39	—	—
Eosinophils (4) [0.18 × 109/l]	≤Median	81	Ref.	—	—	—
	>Median	76	0.80 (0.57-1.13)	0.20	—	—
Thrombocytes (1) [278 × 109/l]	≤Median	82	Ref.	—	—	—
	>Median	78	1.48 (1.05-2.09)	0.02	1.10 (0.75-1.63)	0.62
LDH (8)	Normal	146	Ref.	—	Ref.	—
	>ULN	7	2.94 (1.36-6.33)	0.006	3.74 (1.47-9.54)	0.006
ALP (2)	Normal	152	Ref.	—	—	—
	>ULN	7	2.01 (0.93-4.33)	0.07	—	—
ALAT (3)	Normal	145	Ref.	—	—	—
	>ULN	13	1.55 (0.85-2.80)	0.15	—	—
Albumin (29) (43 g/l)	≤Median	69	Ref.	—	—	—
	>Median	63	0.87 (0.60-1.27)	0.47	—	—

Statistically significant P-values are in bold.

ALAT, alanine aminotransferase; ALP, alkaline phosphatase; BMI, body mass index; CCI, Charlson Comorbidity Index; CI, confidence interval; CTx, chemotherapy; ECF, epirubicin, cisplatin, and fluoropyrimidine; ECOG PS, Eastern Cooperative Group performance status; EOF, epirubicin, oxaliplatin, and fluoropyrimidine; FLOX, 5-flouroupyrimidine and oxaliplatin; FDG-PET, [¹⁸F]2-fluoro-2-deoxy-D-glucose positron emission tomography, GEJ, gastro-esophageal junctional; HR, hazard ratio; LDH, lactate dehydrogenase; LLN, lower limit of normal; NLR, neutrophils to lymphocytes ratio; Ref., reference value (=1.0); UICC, Union for International Cancer Control; ULN, upper limit of normal.

^aNumbers with missing value in parentheses, median values in brackets.

^bMultivariate analysis included age, ECOG PS, CCI, endoluminal stenting, dysphagia, tumor type, T-stage and N-stage. 14 patients with missing variables were excluded.

^cIncludes one patient treated with FLOX.

^dMultivariate analysis included blood tests significant in univariate analysis and endoluminal stenting, dysphagia, tumor type, T-stage, and N-stage. Thirteen patients with missing variables were excluded.

Multivariate analysis of the cohort with available blood tests showed additional poor prognostic value of increased total leucocytes, decreased lymphocytes, and—in 4.6% of patients—elevated lactate dehydrogenase (LDH) (P ≤ 0.004). Thrombocytes and neutrophils to lymphocytes ratio (NLR) was significant in univariate survival analysis (P ≤ 0.02), but not in multivariate (P ≥ 0.62). Kaplan–Meier plots according to OS are shown in Figure 1.

Figure 1.

Kaplan–Meier plots of prognostic factors for overall survival from start date of chemotherapy in patients with resectable gastro-esophageal cancer according to (A) T-stage, (B) N-stage, (C) dysphagia, (D) Charlson Comorbidity Index, (E) Eastern Cooperative Oncology Group (ECOG) performance status, (F) lymphocyte count, (G) neutrophils to lymphocytes ratio, and (H) total leucocyte count. Note: P values are the result of log-rank tests.

Prognostic factors after radical resection

A total of 202 patients (70.9%) achieved radical surgery after preoperative chemotherapy. Median OS from date of radical resection was 3.5 years (IQR 1.5-12.8 years), and OS at 2, 5 and 10 years was 63.4%, 43.1% and 32.2%, respectively. Median recurrence-free survival (RFS) was 2.9 years (IQR 1.1-12.8 years). A total of 142 patients (70.3%) started on postoperative chemotherapy, and 46.8% of these completed all three planned series.

Postsurgical variables and results of prognostic analysis are shown in Table 2. Increasing postneoadjuvant pathological (yp) T-, N- and UICC stage, >10% weight loss, impaired ECOG PS, and age >70 years predicted poor OS from the date of surgery (P ≤ 0.04), while complete downstaging (pCR) was a favorable prognostic factor (P = 0.02). Kaplan–Meier plots according to OS are shown in Figure 2. The same factors were prognostic for RFS (P ≤ 0.04), except for high age and weight loss, which were only marginally significant (P ≤ 0.08). Patients having dose reductions of preoperative chemotherapy tended to have shorter RFS (P = 0.08), and patients not starting postoperative chemotherapy tended to have worse OS (P = 0.07). Most frequent reasons for not starting postoperative chemotherapy were poor PS or postsurgical medical complications (33.3%), surgical complications (23.3%), patient wish (15.0%), intolerance to preoperative chemotherapy (11.7%) and early progression or death (10.0%).

Table 2.

Postsurgical variables and results of survival analysis of 202 patients radically resected after preoperative chemotherapy for gastro-esophageal adenocarcinoma

			Overall survival				Recurrence-free survival
			Univariate analysis		Multivariate analysisa		Univariate analysis		Multivariate analysisa
Variableb	Cut point	Patients (n)	HR (95% CI)	P value	HR (95% CI)	P value	HR (95% CI)	P value	HR (95% CI)	P value
Sex	Female	41	Ref.	—	—	—	Ref.	—	—	—
	Male	161	1.10 (0.73-1.66)	0.64	—	—	1.12 (0.75-1.69)	0.57	—	—
Age, years	<60	70	Ref.	—	Ref.	—	Ref.	—	Ref.	—
	60-69	81	0.96 (0.65-1.42)	0.85	0.98 (0.65-1.47)	0.93	0.96 (0.65-1.42)	0.85	0.99 (0.66-1.49)	0.98
	70+	51	1.57 (1.04-2.36)	0.03	1.47 (0.96-2.25)	0.07	1.48 (0.98-2.24)	0.059	1.42 (0.93-2.17)	0.11
ECOG PS (71)	0	64	Ref.	—	—	—	Ref.	—	—	—
	1	46	1.48 (0.95-2.31)	0.09	—	—	1.58 (1.01-2.46)	0.04	—	—
	2/3	19/2	1.79 (1.03-3.14)	0.04	—	—	1.73 (0.99-3.20)	0.055	—	—
BMI (57)	Normal	64	Ref.	—	—	—	Ref.	—	—	—
	Low	5	1.49 (0.59-3.75)	0.40	—	—	1.48 (0.59-3.74)	0.40	—	—
	High	76	0.87 (0.58-1.28)	0.47	—	—	0.89 (0.60-1.32)	0.55	—	—
Weight lossc (61)	≤10% loss	74	Ref.	—	—	—	Ref.	—	—	—
	>10% loss	67	1.51 (1.04-2.19)	0.03	—	—	1.39 (0.96-2.02)	0.08	—	—
Tumor site	Esophagus/GEJ	33/133	Ref.	—	—	—	Ref.	—	—	—
	Stomach	36	0.88 (0.58-1.36)	0.58	—	—	0.84 (0.55-1.30)	0.44	—	—
Tumor type	Adenocarcinoma	199	Ref.	—	—	—	Ref.	—	—	—
	Adenosquamous carcinoma	3	2.14 (0.68-6.75)	0.19	—	—	2.11 (0.67-6.64)	0.20	—	—
ypT-stage (1)	0	20	Ref.	—	—	—	Ref.	—	—	—
	1	25	3.62 (1.34-9.83)	0.01	—	—	3.60 (1.33-9.77)	0.01	—	—
	2	96	5.13 (2.07-12.7)	<0.001	—	—	5.15 (2.07-12.8)	<0.001	—	—
	3	51	8.38 (3.31-21.2)	<0.001	—	—	9.00 (3.55-22.8)	<0.001	—	—
	4	9	13.6 (4.50-41.3)	<0.001	—	—	14.4 (4.76-43.8)	<0.001	—	—
ypN-stage (1)	0	89	Ref.	—	—	—	Ref.	—	—	—
	1	104	2.41 (1.70-3.41)	<0.001	—	—	2.54 (1.80-3.60)	<0.001	—	—
	2	7	3.21 (1.37-7.49)	0.001	—	—	3.36 (1.44-7.85)	0.01	—	—
	3	2	4.92 (1.19-20.8)	0.003	—	—	4.44 (1.07-18.4)	0.04	—	—
ypUICC stage (1)	0 (pCR)	20	0.30 (0.11-0.83)	0.02	NA	NA	0.31 (0.11-0.84)	0.02	NA	NA
	I	25	Ref.	—	Ref.	—	Ref.	—	Ref.	—
	II	102	1.61 (0.94-2.76)	0.08	1.71 (0.68-4.27)	0.25	1.63 (0.95-2.79)	0.08	1.79 (0.72-4.47)	0.21
	III	49	2.86 (1.60-5.11)	<0.001	3.87 (1.50-9.99)	0.005	3.19 (1.79-5.69)	<0.001	3.96 (1.53-10.2)	0.005
	IV	5	3.87 (1.40-10.7)	0.009	1.64 (0.30-9.05)	0.57	3.80 (1.38-10.5)	0.01	1.73 (0.31-9.56)	0.53
Response UICC-staged (4)	None/PD	83/25	Ref.	—	Ref.	—	Ref.	—	Ref.	-
	Partial	70	0.96 (0.68-1.35)	0.81	0.52 (0.34-0.81)	0.004	0.93 (0.66-1.30)	0.65	0.51 (0.33-0.79)	0.003
	Complete	20	0.18 (0.07-0.43)	<0.001	0.11 (0.04-0.29)	<0.001	0.17 (0.07-0.42)	<0.001	0.11 (0.04-0.29)	<0.001
Preoperative CTx regimen	EOFe	102	Ref.	—	—	—	Ref.	—	—	—
	ECF	100	1.14 (0.83-1.58)	0.42	—	—	1.13 (0.82-1.56)	0.47	—	—
Reduced preoperative CTx dosef (1)	No	129	Ref	—	—	—	Ref.	—	—	—
	Yes	72	1.29 (0.93-1.80)	0.13	—	—	1.35 (0.97-1.88)	0.08	—	—
Start on postoperative CTx	Yes	142	Ref.	—	—	—	Ref.	—	—	—
	No	60	1.38 (0.97-1.95)	0.07	—	—	1.29 (0.91-1.83)	0.15	—	—

Statistically significant P-values are in bold.

BMI, body mass index; CI, confidence interval; CTx, chemotherapy; ECF, epirubicin, cisplatin, and fluoropyrimidine; ECOG PS, Eastern Cooperative Group performance status; EOF, epirubicin, oxaliplatin and fluoropyrimidine; FLOX, 5-flouroupyrimidine and oxaliplatin; GEJ, gastro-esophageal junctional; HR, hazard ratio; NA, not analyzed; pCR, pathological complete response; PD, progressive disease; Ref., reference value (=1.0); UICC, Union for International Cancer Control; yp, post-neoadjuvant pathological.

^aMultivariate analysis included age, ypUICC stage, and response in UICC stage. Four patients with missing variables excluded.

^bNumbers with missing value in parentheses.

^cDifference between pretreatment and postoperative weight.

^dNumerical difference between stage at start on chemotherapy and postoperative stage.

^eIncludes one patient treated with FLOX.

^fDose reduction, treatment delay, or abortion.

Figure 2.

Kaplan–Meier plots of prognostic factors for overall survival from date of surgery for patients radically resected after perioperative chemotherapy according to (A) age, (B) Eastern Cooperative Oncology Group (ECOG) performance status, (C) weight loss, (D) ypT-stage, (E) ypN-stage, (F) yp-UICC stage, (G) response in UICC stage, (H) alanine aminotransferase. Note: P values are the result of log-rank test. pCR, pathological complete response; PD, progressive disease; UICC, Union for International Cancer Control; yp, postneoadjuvant pathological.

Multivariate analysis showed a favorable outcome associated with complete (P < 0.001) and partial (P ≤ 0.003) downstaging, while poor outcome was associated with ypUICC stage III (P = 0.005). A strong correlation between frequency of downstaging and clinical stage at initiation of chemotherapy was observed as shown in Table 3. Downstaging was achieved in 100% of stage IV, 69.2% of stage III, 16.3% of stage II and 10% of stage I tumors (P < 0.001).

Table 3.

Frequency of downstaging after preoperative chemotherapy according to clinical UICC stage^a

UICC stage	Downstaging			Total number
	None n (%)	Partial n (%)	Complete (pCR) n (%)
I	9 (90.0)	0	1 (10.0)	10
II	67 (83.8)	8 (10.0)	5 (6.3)	80
III	32 (30.8)	58 (55.8)	14 (13.5)	104
IV	0	4 (100)	0	4
All stages	108 (54.5)	70 (35.4)	20 (10.1)	198

pCR, pathological complete response; UICC, Union for International Cancer Control.

^aFour patients with missing variables excluded. Downstaging assessed by the numerical difference between stage at start on chemotherapy and postoperative stage. The distribution was significantly skewed (P < 0.001).

Postsurgical blood tests were available in a subset of 42.6%-65.3% of patients. Univariate analysis of blood tests, shown in Supplementary Table S2, available at https://doi.org/10.1016/j.esmogo.2025.100183, showed poor prognosis associated with elevated alanine aminotransferase (ALAT) (in 7.7% of patients) [hazard ratio (HR) 3.00, 95% CI 1.43-6.29, P = 0.004], while white blood cell counts or thrombocytes were not prognostic in this cohort (P ≥ 0.29).

A landmark analysis of residual RFS at 5 years after radical resection (Supplementary Table S3, available at https://doi.org/10.1016/j.esmogo.2025.100183) included 83 recurrence-free survivors. Of these, only three (3.6%) later experienced relapses while 26 (31.3%) died. Although no factor achieved statistical significance, ypT-stage 0 was seemingly associated with a favorable residual RFS (HR 0.34, 95% CI 0.10-1.13, P = 0.08), while age >70 years tended to predict poor prognosis (HR 4.27, 95% CI 1.00-18.3, P = 0.05).

Second cancers

Excluding cutaneous nonmelanoma carcinomas, 10 (3.5%) patients were diagnosed with a synchronous cancer, primarily non-small-cell lung and prostate cancers. During treatment and follow-up, an additional 19 (7.5%) patients were diagnosed with a second cancer, while 4 were diagnosed with a third cancer, totaling 33 cancer diagnoses among 285 patients. Prostate and large bowel were the most common metachronous cancer primaries (Supplementary Table S4, available at https://doi.org/10.1016/j.esmogo.2025.100183). The median time to occurrence of a first metachronous cancer was 8.5 years (range 0.65-13.5 years), and the incidence of second cancers cumulated to 10.1% at 10 years.

Comorbidities and competing causes of death

The most frequent added comorbidities were cardiovascular and cerebral vascular events, depression, ileus and bone fractures. Details are provided in Supplementary Table S5, available at https://doi.org/10.1016/j.esmogo.2025.100183. Depression and diabetes were mostly registered in the early years after treatment whereas cerebral vascular and cardiovascular events, renal failure, ileus and bone fractures occurred later.

A total of 111 (55.0%) patients among 202 that had undergone radical surgery experienced no recurrence. Half of patients without recurrence died during follow-up with a median time to death of 4.7 years from surgery. In half, causes of death were either nonspecific or could not be determined; however, 8.8% of deaths were due to a second cancer, 12.3% died from cardiac events and/or renal failure, and 3.5% from stroke, while treatment-related deaths accounted for 24.6%. Of the latter, four (2.0% of the radically resected patients) died within 1 month, six died within 1-3 months, and three within 3-5 months after surgery from complications, accounting for a postsurgical mortality of 6.4% (1 death per 16 patients radically operated). One patient (<1%) died from complications to postoperative chemotherapy (sepsis) among 142 patients who initiated chemotherapy after surgery.

Discussion

This national audit presents long-term results of perioperative chemotherapy and resection for patients with gastro-esophageal adenocarcinoma in a Western population. Underscoring the need for assessment of treatment benefit outside clinical studies,²³ the overall outcome results were inferior to those presented in pivotal randomized trials, including a 4%-6% points inferior 5-year OS.^4,5 Compared with patients in these trials, distributions of age, sex and PS scores were quite similar. However, only 17% of patients in our cohort had gastric primaries, compared with 25% in the ACCORD-07 and 75% in the MAGIC trial cohorts.^4,5 This difference might in part explain the 5-year OS difference, as patients with gastric cancer may have better outcome from perioperative chemotherapy and resection than those with GEJ primaries.⁶ However, compared with results of perioperative FLOT in the recent ESOPEC trial that included only GEJ primaries, 5-year OS in the current cohort was highly inferior (18.7% points), likely as a result of a much lower fraction of patients receiving R0 resection (70.9% versus 82.4%).⁹

Information on expected outcome before treatment is important for shared decision making on how to treat. Clinical staging and ECOG PS are universally accepted prognostic factors; both, however, are inaccurate at the individual level.²⁴ As demonstrated in this and similar studies, a number of easily measurable clinicopathological variables may provide additional prognostic information.²⁵ CCI and dysphagia were factors of independent prognostic value in the present study, as well as in others,^26,27 but low body mass index (BMI), GEJ primary, high frailty score, old age and Lauren diffuse-type tumors have also been associated with poor outcome.^20,28,29 Molecular factors such as Programmed death-ligand 1 (PD-L1), claudin and human epidermal growth factor 2 (HER2) expression are currently used for treatment stratification in advanced disease but may also become important in the perioperative setting, depending on results of ongoing randomized trials with targeted agents.¹⁰ Moreover, patients with tumors with deficient mismatch repair (dMMR) or microsatellite instability (MSI) potentially achieve no benefit from perioperative chemotherapy¹⁰; MSI-high being observed in 10%-22% of gastric cancers and ∼7% of esophageal adenocarcinomas.³⁰

Routine blood tests potentially reflecting inflammation, poor nutrition or high tumor burden are predictive of poor outcome in gastro-esophageal cancer.³¹ Several blood-based indices have been suggested, such as the Prognostic Nutritional Index, including albumin and lymphocyte counts,³² and the Glasgow Prognostic Index, including C-reactive protein (CRP) and albumin.³³ Common to these is that they were assessed in populations of uncertain relevance for patients undergoing perioperative chemotherapy. A Dutch registry study of 8465 patients with potentially curable gastro-esophageal cancer proposed a predictive model based on treatment modality, clinicopathological factors, albumin, hemoglobin and LDH, but suffered from a large number of missing variables.²⁴ In the current cohort, in addition to LDH, pretreatment total leucocytes and lymphocyte counts were prognostic in multivariate analysis, whereas the NLR, which has repeatably shown prognostic impact in gastro-esophageal cancer,³⁴ was significant in univariate analysis only. The association of lymphocyte counts with outcome has been shown previously,³² and it has been hypothesized that numbers of circulating lymphocytes may reflect antitumor immunity, being correlated with numbers of infiltrating cytotoxic T cells in tumor tissue.^35,36 As shown in the present study, the prognostic impact of white blood cell counts may be absent in pretreated patients, and prospective studies with rigorous sampling protocols in clinically relevant cohorts are needed to confirm the added value of blood-based parameters.³³

Prognostic factors after resection are useful for decisions on adjuvant therapy, in planning a risk-stratified follow-up program, and in informing patients about expected outcome. However, few studies in cohorts treated by perioperative chemotherapy are available. In a multicenter study of 408 patients with gastric cancer, poor prognostic factors after neoadjuvant chemotherapy and surgery were few cycles of chemotherapy administered, ypN+, diffuse-type histology, <15 resected lymph nodes, and age >60 years.³⁷ In this cohort, ypUICC staging and individual staging factors were very strong prognosticators. Especially, patients resected for ypUICC stage III or IV disease had extremely poor outcome, warranting investigations of postoperative, multimodal oncological treatment regimens that are efficient in the palliative setting.³⁸

We also found significant prognostic impact of response to chemotherapy assessed by restaging. In contrast to seemingly futile assessments by repeated computed tomography scans,³⁹ several studies have shown that restaging by comparing clinical stage at baseline with postsurgical stage after chemotherapy is prognostic^21,39 and may also be predictive of benefit of postoperative chemotherapy.²² In a United States National Cancer Database study of 2382 patients with gastric cancer, response at restaging correlated with improved prognosis, and initiation of postoperative chemotherapy was associated with longer survival in the 31% patients with incomplete response (HR 0.64, 95% CI 0.46-0.91), but not in those with pCR (7%), or refractory disease (62%).²² In this study, we could confirm the added prognostic impact of complete and partial response at restaging; downstaging, however, was highly correlated with initial clinical stage, indicating that the variable is less useful in early stages. Alternatively, prognostic and predictive value have been reported by grading pathological response to chemotherapy.⁴⁰ Interim analyses of several ongoing randomized studies of perioperative chemotherapy +/− immune checkpoint inhibitors have shown significant effect of chemo-immunotherapy combinations on the frequency of pCR and downstaging.^41,42 The finding of a strong prognostic significance of pCR and downstaging in cohort studies indicates that this will likely transform into a survival benefit.

Most patients that experience relapse do so within 1-2 years,⁴³ but few studies have sufficiently assessed risk of relapse in long-term survivors. A Korean single-center study of 500 patients radically resected for gastric adenocarcinoma and treated with adjuvant chemotherapy showed that 10.8% recurred after 5 years and only nonmetastatic stage IV disease was predictive for risk of late recurrence.⁴⁴ In a 5-year landmark analysis of 870 esophagus cancer patients (88% adenocarcinomas) from the United States Mayo Clinic, 201 deaths among 416 patients occurred 5 years after esophagectomy, while only 5.6% experienced late recurrence.⁴⁵ In the current study, we could confirm the sustained high mortality rate but rarity of late relapses after the 5-year landmark with further 7 years of follow-up.

Risk of second cancers is increased in gastro-esophageal cancer patients. In 11 812 patients registered in the United States Surveillance, Epidemiology, and End Results (SEER) database with adenocarcinoma of the esophagus, the risk of a second cancer was increased by 11%, including increased risk of stomach, lung and kidney cancer.⁴⁶ In gastric adenocarcinoma, out of 44 041 patients, 4.3% developed a second primary with a median latency of 3 years and the incidence rate of second cancers was 36% higher than expected. Sites of cancers with excess incidence ratio included lung, pancreas and kidney, while the risk of prostate and breast cancer was reduced.⁴⁷ In our small cohort, the median latency of metachronous cancers was 8.5 years, and the 10-year cumulated incidence of a second malignancy was 10%. A second cancer was, however, a rare cause of mortality.

Late morbidity is poorly studied in gastro-esophageal cancer patients treated by chemotherapy and resection. The surgery by itself may cause chronic malnutrition, eating dysfunction, diarrhea, and weight and bone loss,^42,48,49 but chemotherapy may also cause serious late morbidity including cardiac events and renal impairment.⁵⁰ New comorbidities were similar in the present study to those of other cancer diagnoses, where chemotherapy was associated with both cerebral and cardiovascular comorbidity, as well as a higher risk of mental disorders in the years following treatment.⁴³ In addition, we registered early cases of diabetes and, mostly, late cases of renal failure, ileus and bone fractures.

In this elderly and comorbid population, mortality from competing causes was high. Of note, postsurgical complications accounted for a quarter of deaths in patients without recurrence. In a recent meta-analysis 1-month postoperative mortality rates after esophagectomy ranged from 0.6% to 23.8% in different cohorts, depending highly on surgical volume.⁵¹ Our results that excluded nonradically operated patients showed a favorable 2% 1-month mortality but a 6.4% mortality due to complications within 5 months, indicating that longer observation is needed to sufficiently evaluate surgical mortality. Besides centralization of treatment at high-volume centers, optimized patient selection based on risk scores,⁵² better preconditioning,⁵³ and use of minimally invasive surgical techniques⁵⁴ may result in reduced postoperative mortality in modern cohorts. In this cohort, mortality due to postoperative chemotherapy was <1%; in addition, a recent Swedish population-based study of 613 patients surviving at least 5 years after gastric resection showed no excess mortality from preoperative chemotherapy.⁵⁵

Strengths and limitations

This retrospective analysis adds to limited prior evidence of long-term results in Western patients as most similar studies have been done in Asian populations or in institutional cohorts with shorter follow-up. Obtained by journal audit and being subjective in their assessment, some variables may be difficult to reproduce, such as dysphagia and CCI, as well as added comorbidities and causes of death were ill defined in some cases. We studied pretreatment blood tests only in a subset of patients who deviated according to age from the full cohort but were otherwise similar. Some variables could only be obtained in a fraction of patients, limiting the number of factors included in multivariate analyses. We had no information on prognostic Lauren subtype, tumor grade or histological response, and no molecular factors were available.²⁰ The inclusion of patients with MSI/dMMR due to lack of molecular information may impact results due to potentially reduced efficacy of chemotherapy in this group of patients. In the resected cohort, we chose to focus only on patients radically resected, as R0 resection is necessary for cure,⁵⁶ and results in this subcohort are not intention-to-treat.

Although the standard perioperative treatment regimen has changed and future outcome results will likely be improved, we believe that results of prognostic factors, second cancers and comorbidities in the current cohort are transferable to modern cohorts. This also includes findings that downstaging and pCR are strongly associated with long-term survival. Still, prospective, population-based studies of the long-term outcome of patients treated with perioperative FLOT are warranted.

Conclusions

Outcomes of patients with resectable gastro-esophageal adenocarcinomas treated with perioperative chemotherapy in this national cohort were inferior to those of pivotal randomized trials, but, also, gastric tumors were rarer. A multitude of clinical, pathological, imaging, molecular and blood-based prognostic factors have been disclosed in this and other studies that could improve prediction, and new machine learning tools and artificial intelligence may efficiently combine such factors to better guide treatment selection and prognostication in future patients.²³ Moreover, not all patients benefit from chemotherapy, and assessment of treatment response could become clinically useful to identify patients with no benefit from postoperative chemotherapy.²² Downstaging by postoperative restaging is prognostic; it is, however, strongly correlated with pretreatment clinical stage and less useful in early-stage disease.