When and where: A study on gastric cancer recurrence after neoadjuvant/perioperative chemotherapy

Abstract

BACKGROUND

Neoadjuvant chemotherapy is the standard of care for locally advanced gastric cancer (GC) in the Western countries. However, limited data are available on the timing and patterns of recurrence in this setting.

AIM

To evaluate recurrence patterns and timing of neoadjuvant/perioperative chemotherapy in patients with GC, analyze factors influencing recurrence, update relapse-free survival (RFS) and overall survival (OS) estimates.

METHODS

A total of 91 patients with resectable GC were enrolled between September 2010 and August 2016. Participants were then randomized in a 1:1 ratio to receive either 4 cycles of neoadjuvant chemotherapy (arm A) or a preoperative plus postoperative consisting of 2 cycles of docetaxel, oxaliplatin, and capecitabine chemotherapy, followed by 2 additional cycles after surgery (arm B).

RESULTS

Fifty-two out of 84 patients (61.9%) experienced disease recurrence, and for 41 the site of disease was identified. Distant, peritoneal, and local relapses occurred in 21 (51.2%), 11 (26.8%), and 5 (12.2%) patients, respectively. The median follow-up was 79 months (range: 4.2-120 months). The 5-years RFS was 39.0% [95% confidence interval (CI): 27.9-49.9]. cT4 tumors had a higher risk of recurrence compared to those with cT2-3 tumors (hazard ratio = 1.92; 95%CI: 1.03-3.56; P = 0.038). The 5-years and 7-years OS was 58.7% (95%CI: 41.9-72.2) and 47.3% (95%CI: 29.4-63.2) in arm A and it was 55.0% (95%CI: 38.4-68.9; P = 0.806) and 46.8% (95%CI: 30.7-61.5; P = 0.806) in arm B, respectively.

CONCLUSION

Distant metastases are the most common relapse sites, within two years after surgery; RFS depends on tumor size, histology and treatment response; median OS was 69.6 months with favorable long-term outcomes.

Key Words: Gastric cancer; Neoadjuvant chemotherapy; Perioperative chemotherapy; Time of recurrence; Site of recurrence; Postoperative recurrence

Core Tip: This randomized study provides long-term data on recurrence patterns after neoadjuvant or perioperative chemotherapy for resectable gastric cancer. Most relapses occurred at distant sites, with the highest risk within the first two years after surgery. These findings support risk stratification particularly given by interaction between tumor regression score and lymph nodes involvement. Time-dependent modelling suggested that the intestinal subtype is associated with a decreasing recurrence risk over time; however, this finding requires further validation.

INTRODUCTION

Worldwide, gastric cancer (GC) is among the most common and lethal malignancies, ranking fifth in both incidence and mortality[1]. Curative resection has long been considered the only potential curative treatment for GC. Although neoadjuvant chemotherapy improves cure rate compared with resection alone[2,3], GC still recurs in 25%-40% of patients[4,5].

Neoadjuvant therapy aims to eliminate micro-metastases and achieve tumor downstaging, thereby increasing the rate of surgical resections without residual disease[6-8], and ultimately improving survival. Administering chemotherapy before gastrectomy, allows treatment in patients with better clinical condition and in tumors with optimal vascularization, facilitating assessment of chemosensitivity. However, disease progression during treatment may reflect the development of therapeutic resistance.

Understanding recurrence patterns is crucial for developing and improving treatment strategies and follow-up. The main information on relapse sites comes from studies that included patients who did not receive neoadjuvant chemotherapy[9-11]. In the United States, local-regional recurrences and hematogenous metastases are fairly common[9-11]. Patients in Japan and South Korea experienced lower rates of local regional recurrence compared with those in China (7%-10% vs 32.4%)[12-14].

In light of the limited prospective data on recurrence patterns in Western populations, the present study investigates the timing and sites of recurrence following neoadjuvant or perioperative chemotherapy and updates survival outcomes from the GASTRODOC trial[15].

MATERIALS AND METHODS

The current study was based on the GASTRODOC study, a multicenter randomized open-label phase-II trial. In this superiority study, patients meetings the eligibility criteria were randomized to treatment groups (1:1), receiving either neoadjuvant chemotherapy (arm A) or perioperative approach (arm B). Allocation was based on center-specific randomization lists generated using a permuted-block design to ensure balance within each participating site.

Eligible participants were adults (18-75 years) of either sex with advanced, surgically resectable M0 GC (defined as ≥ T3 or bulky N+, according to the tumor node metastasis (TNM) classification)[16], excluding cancer of the gastric cardia. Staging laparoscopy with peritoneal washing was performed in all cases, and individuals showing positive peritoneal cytology and/or signs of peritoneal involvement were excluded thereafter. Other inclusion criteria included an Eastern Cooperative Oncology Group performance status 0-1 assessed and recorded at baseline, prior to the initiation of treatment, and sufficient bone marrow. The criteria used to define “bulky lymph node metastases” were based on those reported by Yoshikawa et al[17]: “At least one node of ≥ 3 cm in diameter or at least three consecutive nodes of ≥ 1.5 cm diameter in first or second level lymph node stations”.

Patients were excluded in the presence of linitis plastica, ongoing chronic systemic immunotherapy, clinically significant coronary artery disease, and any myocardial infarction or inadequately controlled hypertension within the 12 months preceding treatment. For further details on the inclusion/exclusion criteria, please refer to the original published work[15]. The primary aim of this study was to evaluate the timing and sites of recurrence in patients treated with neoadjuvant/perioperative chemotherapy.

The secondary objective was to explore parameters associated with the recurrence patterns identified in the primary analysis and to provide updated estimates of the relapse-free survival (RFS) and overall survival (OS). RFS was defined as the time elapsed between randomization and either the first documented disease progression or death from any cause. OS was measured from randomization until death from any cause, or last known follow-up.

The study was conducted in compliance with Good Clinical Practice, the Declaration of Helsinki (1964) and its amendments, Directive 2001/20/EEC and applicable national regulation for non-profit clinical studies.

The IRCCS Istituto Romagnolo per lo Studio dei Tumori (IRST) “Dino Amadori” in Meldola (Italy) was responsible for overall study coordination and monitoring activities, both remote and on site. Surgical procedures at participating centers were carried out by surgeons belonging to the Italian Research Group for GC. The trial was registered on ClinicalTrial.gov website (No. NCT01876927) and in the EudraCT database (No. 2010-020189-37).

Treatment

Patients assigned to arm A received 4 cycles of chemotherapy administered prior to surgery, whereas patients in arm B received 2 cycles before surgery and two additional cycles after surgery, unless disease progression, unacceptable toxicity, or withdrawal of consent occurred. Surgical resection was scheduled within 3-6 weeks after the completion of the final preoperative cycle (cycle 4 in arm A and cycle 2 in arm B). The chemotherapy regimen included docetaxel 35 mg/m² given on days 1 and 8 as 1-hour infusion, oxaliplatin 80 mg/m² on day 1 over 2-hour infusion, and capecitabine 750 mg/m² given orally twice daily for 14 days, with cycles repeated every three weeks. Surgery consisted of complete tumor removal with at least a D2 lymphadenectomy, in accordance with Japanese GC Association guidelines[18]. Subtotal gastrectomy was performed for distal cancer when appropriate. Splenectomy was reserved for patients with GC involving or adherent to the spleen, as well as for patients showing intraoperative macroscopic lymph nodes disease at the splenic hilum or along the distal splenic artery.

Response assessment

Response to treatment was evaluated by computed tomography (CT) imaging performed after every two cycles in both arms. Histopathological response was evaluated according to the Becker criteria[19], which classify regression based on the estimated percentage of vital tumor cells within the macroscopically defined tumor bed. The categories were defined as follows: (1) Complete pathological response; (2) Subtotal regression with less than 10% residual tumor cells; (3) Partial regression with 10%-50% residual tumor cells; and (4) Minimal or absent regression with more than 50% residual tumor cells. Tumor stage and margin status (R0-R2) were assessed at each participating center by local pathologists following the TNM classification of malignant tumors[16].

To evaluate the primary objective of the study, we classified patients into two groups according to their tumor regression score and lymph node status (ypN+): (1) Good response group, consisting of patients with ypN0 status and either complete (Becker 1a), near-complete (Becker 1b) or partial tumor regression (Becker 2); and (2) Poor response group, comprising patients with ypN1/2/3 status with Becker 1a, 1b or 2 regression, or those with Becker 3 regression and ypN0/N1/N2/N3 status.

Follow-up began 6 months after randomization and continued until disease progression. It consisted of clinical evaluations and laboratory tests every 3 months during the first two years and every six months thereafter. Radiological surveillance (chest or abdominal imaging) was carried out at six months intervals for up to five years, or until progression of disease. In patients who underwent subtotal gastrectomy, gastroscopy was carried out at one year after randomization and then annually, while in those who underwent total gastrectomy it was performed every 2 years, for a maximum of 5 years.

Statistical analysis

Time-to-event outcomes (OS and RFS) were analyzed using the Kaplan-Meier curves, and compared with the log-rank test. The complement of the Kaplan-Meier curve was used to depict the cumulative incidence. Survival rates were reported with corresponding 95% confidence intervals (CIs). Cox proportional hazards regression models were applied to evaluate the association between clinical, pathological variables and time to event outcomes. Variables showing a P value < 0.05 in univariable analyses were entered in the multivariable models. The proportional hazards assumption was evaluated for each covariate and for the overall model using Schoenfeld residuals; when this assumption was not met, variables were modelled as time-dependent covariates as appropriate. Hazard ratios (HRs) with 95%CIs and associated P values were reported to quantify the strength of association between covariates and time to event endpoints. A P value < 0.05 was considered statistically significant. All statistical analyses were conducted using Stata/SE version 15.1 for Windows (StataCorpLP, College Station, TX, United States).

RESULTS

A total of 106 patients were recruited across 13 centers between September 21, 2010, and August 19, 2016, with 55 patients assigned to arm A and 51 to arm B. Of these, 91 patients were deemed eligible for the GASTRODOC study, which evaluated compliance in both arms (A and B), with 46 patients in arm A and 45 in arm B. Eighty-four patients were evaluable for disease recurrence. Among all enrolled patients, 61.9% experienced recurrence, with distant metastasis being the most common site of recurrence (median follow-up of 79 months, range: 4.2-120 months). The site of recurrence was identified in 41 patients, whereas 8 patients died without evidence of recurrence and, in three cases, the site of recurrence remained unknown (Figure 1). Demographics, clinicopathological characteristics, operative and pathological findings of the study cohort are shown in Table 1.

Figure 1 Allocation of patients. A: In the GASTRODOC trial; B: The long-term results. AE: Adverse event; RFS: Relapse-free survival.

Table 1 Baseline patient characteristics (n = 91), n (%).

	A: 4 DOC + surgery (n = 46)	B: 2 DOC + surgery + 2 DOC (n = 45)	Overall
Median age, years (range)	63 (39-74)	66 (33-75)	64 (33-75)
Gender
Male	32 (69.6)	32 (71.1)	64 (70.3)
Female	14 (30.4)	13 (28.9)	27 (29.7)
ECOG performance status
0	42 (91.3)	43 (95.6)	85 (93.4)
1	4 (8.7)	2 (4.4)	6 (6.6)
cTN
T2N+	2 (4.3)	1 (2.2)	3 (3.3)
T3N0	8 (17.4)	5 (11.1)	13 (14.2)
T3N+	18 (39.1)	16 (35.6)	34 (37.4)
T4aN0	4 (8.7)	3 (6.7)	7 (7.7)
T4aN+	13 (28.3)	18 (40.0)	31 (34.1)
T4bN+	1 (2.2)	2 (4.4)	3 (3.3)
cT
T2	2 (4.3)	1 (2.2)	3 (3.3)
T3	26 (56.5)	21 (46.7)	47 (51.6)
T4a	17 (37.0)	21 (46.7)	38 (41.8)
T4b	1 (2.2)	2 (4.4)	3 (3.3)
cN
N0	12 (26.1)	8 (17.8)	20 (22.0)
N1	14 (30.4)	16 (35.6)	30 (33.0)
N2	4 (8.7)	7 (15.6)	11 (12.0)
N3	3 (6.5)	3 (6.7)	6 (6.6)
N+	13 (28.3)	11 (24.3)	24 (26.4)
Response to treatment (Becker + lymph nodes)
Poor response	27 (62.8)	32 (78.1)	59 (70.2)
Good response	16 (37.2)	9 (21.9)	25 (29.8)
Unknown	3	4	7
Lauren histotype
Diffuse	22 (50.0)	13 (29.5)	35 (39.8)
Intestinal	18 (40.9)	23 (52.3)	41 (46.6)
Other	4 (9.1)	8 (18.2)	12 (13.6)
Unknown	2	1	3

ECOG: Eastern Cooperative Oncology Group; DOC: Docetaxel, oxaliplatin, capecitabine.

Relapses occurred most frequently within the first two years of follow-up. In particular, 16 patients (39.0%) relapsed during the first year and 11 patients (26.8%) during the second year. Distant metastases were the most common site of relapse, occurring in 21 patients (51.2%), while the peritoneum represented the second most frequent site, observed in 11 patients (26.8%).

Table 2 summarizes the sites of first recurrence documented during the early follow-up period. Figure 2 illustrates the Kaplan-Meier curve showing cumulative recurrence by disease location; patients presenting with multiple sites of disease were counted more than once.

Figure 2  Inverse Kaplan-Meier representing cumulative proportion of patients with disease recurrence (n = 41).

Table 2 Sites of first recurrence during the first seven years of follow-up (n = 41).

Site of recurrence	Time
	< 1 year	1-2 years	2-3 years	3-4 years	4-5 years	7-8 years	Total
Peritoneal metastases or carci/carcinosis (n)	6	0	3	1	1	0	11
Column (%)	37.5	0.0	42.9	33.3	33.3	0.0	26.8
Local metastases (n)	2	3	0	0	0	0	5
Column (%)	12.5	27.3	0.0	0.00	0.0	0.0	12.2
Distant metastases (n)	6	8	3	1	2	1	21
Column (%)	37.5	72.7	42.9	33.3	66.7	100.0	51.2
Multiple metastases (n)	2	0	1	1	0	0	4
Column (%)	12.5	0.0	14.2	33.34	0.0	0.0	9.8
Total (n)	16	11	7	3	3	1	41
Row (%)	39.0	26.8	17.1	7.3	7.3	2.5	100.00

Demographic and clinical factors, as well as inflammatory indices and their potential association with disease recurrence at one year, were analyzed; however, none of these factors reached statistical significance. In the cohort of 84 patients from the GASTRODOC study who were eligible for RFS analysis, the estimated 5-year RFS rate was 39.0% (95%CI: 27.9-49.9). Figure 3A shows the RFS results by treatment arm in the GASTRODOC study, with no apparent long-term differences observed between arms (P = 0.235).

Figure 3 Kaplan-Meier curves. A: Representing relapse-free survival by arm and 95% confidence interval (n = 84); B: Representing overall survival by arm and 95%confidence interval (n = 86).

In the univariable analysis, patients with the intestinal Lauren subtype demonstrated significantly better RFS (5-year RFS = 61.7%, 95%CI: 44.0-75.3) compared with those with the diffuse histotype (5-year RFS = 18.4%, 95%CI: 6.9-34.3, P < 0.001). Similarly, patients who exhibited a good pathological response to therapy had superior RFS outcome (5-year RFS = 72.6%, 95%CI: 48.6-86.7) compared with those with poor response (5-year RFS = 30.4%, 95%CI: 18.4-43.3, P < 0.017). Patients with cT4 tumor had a lower 5-year RFS (5-year RFS = 28.3%, 95%CI: 14.6-43.8) compared to patients with cT3 (5-year RFS = 48.8%, 95%CI: 32.5-63.3, P = 0.051).

In Table 3, the results of the multivariable Cox regression model are reported. The proportional hazards assumption for each covariate was assessed using Schoenfeld residuals, which indicated that the histological subtype violated the proportional hazards assumption. As a result, histotype was modelled as a time-dependent covariate to account for non-proportionality over the follow-up period. Although clinical T stage (cT) stage did not reach statistical significance in the univariable analysis, it was included in the multivariable model because of its established prognostic relevance.

Table 3 Multivariable analysis for relapse-free survival.

	Multivariable model
	HR (95%CI)	P value
cT
2-3	1.00
4	1.92 (1.03-3.56)	0.038
Response to treatment (Becker + lymph nodes)
Poor response	1.00
Good response	0.36 (0.15-0.85)	0.020
Lauren histotype (main effect)
Diffuse	1.00
Intestinal	1.26 (0.44-3.64)	0.665
Other	0.54 (0.15-1.90)	0.335
Lauren histotype (time-dependent effect)
Diffuse	1.00
Intestinal	0.95 (0.91-0.99)	0.034
Other	1.01 (0.98-1.04)	0.462

HR: Hazard ratio; CI: Confidence interval.

In the multivariable Cox regression analysis, cT, treatment response, and Lauren histotype (modelled as a time-dependent covariate) were evaluated for their association with RFS. Patients with cT4 tumors had a significantly higher risk of recurrence than those with cT2-3 tumors (HR = 1.92; 95%CI: 1.03-3.56; P = 0.038). A good response to treatment was independently associated with a reduced risk of recurrence (HR = 0.36; 95%CI: 0.15-0.85; P = 0.020).

The main effect of Lauren histotype did not show a significant association with RFS (intestinal vs diffuse: HR = 1.26; 95%CI: 0.44-3.64; P = 0.665; other vs diffuse: HR = 0.54; 95%CI: 0.15-1.90; P = 0.335). However, because of violation of the proportional hazards assumption, histotype was further modelled as a time-dependent covariate. In this extended model, the time-dependent effect for the intestinal subtype was statistically significant (HR per unit time = 0.95; 95%CI: 0.91-0.99; P = 0.034), indicating that the risk associated with this histotype decreased over time.

In the GASTRODOC cohort, 86 patients were eligible for OS analysis; the median OS was 69.6 months (95%CI: 44.2-115.1). In our study, the 5-year and 7-year OS rates were 58.7% (95%CI: 41.9-72.2) and 47.3% (95%CI: 29.4-63.2) in arm A, and 55.0% (95%CI: 38.4-68.9) and 46.8% (95%CI: 30.7-61.5) (P = 0.806) in arm B, respectively (Figure 3B). As shown in Table 4, in univariable analysis for OS, an increased risk of death was associated with body mass index < 25, cT4 stage (vs cT3), diffuse Lauren subtype (vs intestinal), and a poor pathological response to therapy (vs good response). These factors are not statistically significant in the multivariable model.

Table 4 Univariable overall survival.

Variable	Number of patients	Number of events	5 years OS (95%CI)	Median OS in months (95%CI)	P value (log-rank test)
Overall patients	86	45	57.1 (45.5-67.1)	69.6 (44.2-115.1)
BMI
Underweight or normal (< 25)	32	24	42.9 (25.5-59.1)	41.6 (20.7-78.3)	0.008
Overweight or obesity (≥ 25.0)	53	21	64.8 (49.4-76.6)	115.1 (59.9-NE)
Unknown	1	0
cT
T2	3	2
T3	45	18	63.4 (47.2-75.8)	NR	0.045
T4	38	25	48.0 (30.5-63.5)	59.9 (39.1-89.4)
Lauren histotype
Diffuse	34	24	37.8 (21.3-54.3)	40.6 (21.3-69.6)	0.001
Intestinal	38	13	72.7 (54.5-84.1)	NR
Other	14	7
Unknown	2	1
Treatment
Arm A	44	21	58.7 (41.9-72.2)	78.2 (40.5-NE)	0.806
Arm B	42	24	55.0 (38.4-68.9)	65.0 (41.1-115.1)
Response (Becker + lymph nodes)
Poor responders	57	32	50.4 (36.4-62.8)	60.5 (40.5-NE)	0.034
Good responders	24	8	85.7 (61.7-95.2)	115.1 (65.0-NE)

Strata with fewer than 15 patients were excluded from the log-rank test calculation. OS: Overall survival; CI: Confidence interval; BMI: Body mass index; NE: Not estimable from statistical software; NR: Median not reached.

DISCUSSION

Knowing the site, the timing and risk factors of recurrence after curative resection of GC could improve the strategy of postoperative treatment and follow-up planning. Differences in recurrence patterns between Eastern Asian (e.g., Japan and South Korea) and Western countries likely reflect a complex interaction of stage at diagnosis, tumor biology, surgical quality, and therapeutic approaches. Earlier diagnosis and standardized D2 surgery in East Asia may favor improved locoregional control, while more advanced and biologically aggressive disease in Western populations may predispose to earlier systemic and peritoneal recurrence.

It is important to acknowledge that most pivotal studies (MAGIC, FNCLCC, FLOT4)[6-8] on perioperative treatment for GC were not designed to systematically analyze patterns of failure, and therefore provide only limited or indirect information on the anatomical distribution of recurrences.

In this study, we observed that 39.0% of relapses occurred within the first year after gastrectomy. This result is interesting when compared to 71.8% and 46.3% rates, reported by a Dutch trial[4] and a Memorial Sloan Kettering Cancer Center study (MSKCC)[20], respectively. All studies agree that the majority of relapses occurred within the first two years. In particular, the relapse rate within 2 years after gastrectomy was 84% in Nakauchi et al[20], whereas it was 65.8% in our study. A retrospective review of a prospectively maintained database from the University of Texas MD Anderson Cancer Center reported different results, with 125 patients (42.8%) experiencing relapse after preoperative treatment during the follow-up[5]. Differences in tumor characteristics, patient populations, and perioperative treatments probably result in variability in survival rates among studies.

Across these three Western trials, distant metastases dominated as the first site of recurrence in 26.4% patients in the Dutch trial[4], peritoneal metastases in 48.8%[5] and 45.6% patients[20] in the other two studies. In the GASTRODOC study, the most frequent sites of relapse were distant in 51.22% of the patients, and only 26.83% of peritoneal relapse. This could be related to the fact that we performed laparoscopy before starting chemotherapy, whereas the above studies did not always describe the use of this staging technique. The reduction of local and/or peritoneal recurrences must also be correlated with performing gastrectomy in high-volume surgeries. In our study, 9.7% of patients had multiple metastatic sites, compared to 39.2% reported by Mokadem et al[4]. Once the relapse sites and the time of reappearance have been identified, it is helpful investigate the causes of the relapse.

The two Western studies mentioned above found a higher risk of recurrence in patients with a ypN+ stage disease[4,20], partial or no tumor regression[4], ypT4[20], or those receiving 3 than six chemotherapy cycles. Pathologic response to neoadjuvant chemotherapy is typically assessed based only on the primary cancer. In the GASTRODOC study, the risk of relapse was related to poor response to therapy and cT4 stage, while the intestinal subtype showed a decreasing risk over time. In this trial, we considered the degree of response to neoadjuvant chemotherapy as a combination of pathological response to the primary tumor and ypN+. The same evaluation of response has been applied by Shao et al[21]. The combination of tumor regression grade (TRG) with ypN status may yield a more comprehensive and reproducible evaluation method for predicting disease control[22,23]. Although TRG after neoadjuvant chemotherapy reflects the response of the primary tumor, it does not fully capture residual systemic disease. ypN remains one of the strongest prognostic factors in GC. Combining TRG with ypN provides a more comprehensive assessment, enabling stratification into groups with distinct recurrence risk and survival outcomes. Several analyses, including the retrospective study by Yin et al[24], have shown that this combined approach has superior prognostic value compared with TRG alone, supporting its use for post-operative risk assessment and potential tailoring of adjuvant therapy. Similarly, a post-hoc analysis of the MAGIC trial by Smyth et al[25] confirmed that, although both TRG and ypN were associated with survival in univariate analysis, only ypN remained independently predictive in multivariate models, highlighting the critical prognostic weight of residual nodal disease after neoadjuvant therapy.

Existing evidence on pathological complete response (pCR) in locally advanced GC is derived mainly from studies with limited sample sizes and heterogeneous populations, including patients with residual ypN+, which may limit the generalizability of the reported outcomes[26-28].

Patients with resectable GC who achieve pCR after neoadjuvant chemotherapy generally experience better outcomes than patients without pCR[29]. In our study, however, 2 out of 3 patients with Becker 1a had disease relapse, suggesting that complete remission is not an absolute certainty of recovery.

Limited attention has been given to patients with poor response to neoadjuvant chemotherapy. A large Chinese multicenter retrospective study reported that in patients with poor tumor regression after neoadjuvant therapy, the peak recurrence occurs within 1 year after surgery, with distant metastases (26.1%) and peritoneal metastases (21.1%) as the primary recurrence patterns[30]. Another study[31] similarly showed that patients with locally advanced GC with less than pCR to neoadjuvant chemotherapy had poor outcomes due to the high rates of distant metastases.

While poor pathological response to neoadjuvant chemotherapy is associated with aggressive disease and a higher likelihood of relapse, current evidence does not support omitting adjuvant therapy. Accordingly, completion of perioperative chemotherapy is generally advised, although modification of the adjuvant regimen in poor responders remains investigational and unsupported by prospective evidence.

Although neoadjuvant chemotherapy offers survival benefits, the results remain modest. The MAGIC[6] and FNCLCC[7] trials reported 5-year OS rates of 36.3% and 38%, respectively. In the FLOT 4 trial[8], 5-year OS was 45% in the experimental arm. The two treatment arms of GASTRODOC, neoadjuvant and perioperative, also demonstrated comparable 7-year survival rates. As a phase II study, this trial was not powered to detect survival differences between neoadjuvant and perioperative strategies. Therefore, the absence of statistically significant differences should be interpreted as a reflection of the study’s limited statistical power.

Compared with previous trials (MAGIC, FNCLCC, FLOT, the Dutch trial, and MSKCC)[4,6-8,20], the GASTRODOC study offers longer follow-up and integrates pathological tumor response with ypN+. It also demonstrates time-dependent recurrence patterns for the intestinal Lauren histotype, providing novel insights into disease dynamics. Molecular biomarkers such as human epidermal growth factor receptor 2 (HER2), microsatellite instability (MSI), and programmed death-ligand 1 (PD-L1)/combined positive score (CPS) may influence recurrence patterns in GC. HER2 positivity has been associated with a higher likelihood of distant metastases, particularly hepatic, rather than peritoneal or loco-regional relapse[32]. In contrast, MSI-high (MSI-H) tumors are generally associated with a more favorable prognosis, including improved OS compared with microsatellite stable tumors, despite a lower pathological response to neoadjuvant chemotherapy[33]. High PD-L1/CPS may further reflect an immunologically active tumor microenvironment that could modulate the pattern of recurrence, potentially influencing systemic vs peritoneal relapse. While distinct recurrence patterns by MSI or PD-L1/CPS status remain incompletely defined, prospective evaluation of site-specific relapse according to these biomarkers is warranted.

The intestinal Lauren subtype is more frequently MSI-H and HER2-positive, with higher PD-L1 expression in MSI-H cases, suggesting a biologically distinct subgroup that may exhibit unique recurrence patterns and potential responsiveness to targeted or immunotherapy approaches[34].

The combination of anti-PD-L1 therapy with chemotherapy has shown encouraging outcomes in the perioperative treatment GC. Results from the phase III MATTERHORN study[35] showed significantly improved event-free survival (i.e., a reduction in recurrence or death) with the addition of durvalumab to FLOT compared to FLOT alone. After a median follow-up of 31.5 months, the 2-year event free survival was 67.4% among the participants in the durvalumab group and 58.5% in the placebo group (HR for event or death = 0.71; 95%CI: 0.58-0.86; P < 0.001)[35].

Although several variables were significantly associated with outcomes in univariable analysis, these associations did not remain statistically significant in the multivariable model. In the GASTRODOC trial, the sites of relapse did not appear to influence survival (Figure 2). This result is in agreement with that of Ministrini et al[36], who reported worse survival in cases with multiple metastases.

Only 7.3% of patients experienced relapse within the third and fourth years after surgery, and just 2.4% relapsed beyond the fifth year, mostly with distant metastases. Twenty out of 27 patients were classified as poor responders based on TRG and ypN+, each underwent abdominal and chest CT every six months. These results indicate that the first two years after surgery are the most critical for surveillance. The follow-up strategies in GC are not yet standardized, and the optimal schedule remains controversial[18,37-39].

This study has several limitations. Firstly, GASTRODOC was a small study: The relatively small cohort limits the robustness of subgroup analyses and should be interpreted as exploratory and hypothesis-testing results. Second, we could not analyze biological factors such as MSI and CPS, and we did not use immunotherapy, which represents a new frontier in neoadjuvant chemotherapy. The third limitation is that our study considered two arms: Neoadjuvant chemotherapy and perioperative chemotherapy. No additional survival benefits seem to be conferred by adjuvant chemotherapy after a pCR[40] or in poor responders[31]. Nevertheless, despite a poor response to neoadjuvant chemotherapy, completing of at least four cycles of adjuvant chemotherapy may offer survival benefits[30].

This study also has different strengths: (1) It was a prospective trial; (2) It considered a single type of chemotherapy even if administered with two modalities; (3) It had a long observation period; and (4) It had a careful radiological staging accompanied by laparoscopy and surgery in high-volume centers.

CONCLUSION

In conclusion, the GASTRODOC trial found that recurrences within the first two years after gastrectomy were most common, and that RFS was influenced by tumor size, histology, and response to treatment; these findings should be considered exploratory and hypothesis-generating. The docetaxel, oxaliplatin, capecitabine regimen is an interesting option that could be further developed, but it is necessary to improve the response to neoadjuvant treatment, for example, combining chemotherapy with immunotherapy may help to reduce the risk of relapse, especially distant metastases.

ACKNOWLEDGEMENTS

The authors thank the Ministry of Health who partly supported this study.