Triplet vs doublet adjuvant chemotherapy for gastric cancer: A systematic review and meta-analysis

Abstract

BACKGROUND

Gastric cancer (GC) remains a major cause of cancer-related mortality worldwide. For patients undergoing curative gastrectomy, postoperative adjuvant chemotherapy is an established strategy to eradicate micrometastatic disease, reduce recurrence, and improve long-term survival. Platinum-fluoropyrimidine doublet regimens are widely used; however, the potential benefit and safety of intensified taxane-platinum-fluoropyrimidine triplet regimens in the adjuvant setting remain uncertain.

AIM

To systematically compare the efficacy and safety of taxane-platinum-fluoropyrimidine triplet chemotherapy vs platinum-fluoropyrimidine doublet chemotherapy as postoperative adjuvant treatment for GC.

METHODS

PubMed, EMBASE, and the Cochrane Library were searched from inception to December 31, 2025. Randomized controlled trials (RCTs) comparing taxane-platinum-fluoropyrimidine triplet chemotherapy with platinum-fluoropyrimidine doublet chemotherapy in patients receiving postoperative adjuvant treatment after curative gastrectomy for histopathologically confirmed GC were included. The primary outcomes were overall survival (OS) and progression-free survival (PFS). Secondary outcomes included objective response rate (ORR), disease control rate (DCR), and grade 3-4 adverse events (AEs) when reported. Study quality was assessed using the Cochrane Risk of Bias tool. Meta-analysis was performed using RevMan 5.4. Random-effects models were applied when substantial heterogeneity was present, defined as I² ≥ 50% or P < 0.10. Risk ratios (RRs) were calculated for dichotomous outcomes, and hazard ratios were calculated for survival outcomes, both with 95% confidence intervals (95%CIs).

RESULTS

Twelve RCTs involving 11126 patients were included. Triplet chemotherapy was associated with a significantly higher ORR than doublet chemotherapy (RR = 1.73, 95%CI: 1.64-1.82, P < 0.0001), with substantial heterogeneity (I² = 84%). Triplet chemotherapy also achieved a higher DCR (RR = 1.22, 95%CI: 1.15-1.28, P < 0.0001; I² = 61%). However, neither OS [hazard ratio (HR) = 1.24, 95%CI: 0.90-1.72, P = 0.19; I² = 96%] nor PFS (HR = 1.52, 95%CI: 0.93-2.50, P = 0.09; I² = 94%) differed significantly between treatment groups. Grade 3-4 AE events, including leukopenia, neutropenia, anemia, thrombocytopenia, nausea, vomiting, diarrhea, fatigue, and anorexia, were not significantly different between triplet and doublet regimens. Publication bias was suggested for ORR and DCR.

CONCLUSION

In this meta-analysis, taxane-platinum-fluoropyrimidine triplet chemotherapy did not demonstrate a clear OS or PFS advantage over platinum-fluoropyrimidine doublet chemotherapy in the postoperative adjuvant setting. Although intensified regimens may be feasible in selected high-risk patients, treatment selection should be individualized according to pathological stage, recurrence risk, postoperative recovery, performance status, and tolerability.

Key Words: Gastric cancer; Adjuvant chemotherapy; Postoperative chemotherapy; Triplet chemotherapy; Doublet chemotherapy; Meta-analysis

Core Tip: This comprehensive meta-analysis of 12 randomized controlled trials encompassing 11126 patients demonstrates that taxane-platinum-fluoropyrimidine triplet chemotherapy significantly improves objective response rate (1.73-fold increase) and disease control rate (1.22-fold increase) compared to platinum-fluoropyrimidine doublet regimens in neoadjuvant treatment of gastric cancer. Despite superior tumor response, no significant survival benefits were observed. Importantly, safety profiles were comparable between regimens. These findings support individualized treatment selection, with triplet regimens preferred for maximal tumor downstaging when surgical resectability is borderline.

INTRODUCTION

The global burden of gastric cancer (GC) remains high, as it is the fifth most common malignancy and the third leading cause of cancer deaths worldwide[1], with approximately 769000 annual deaths and 1.09 million new cases per year. GC is one of the most common cancers worldwide, and despite decreasing incidence in Western countries, it remains highly endemic in East Asia, predominantly China, Japan and South Korea contributing to > 60% of global burden[2]. Although the prognosis of GC has improved recently, overall 5-year survival rate in most countries is only about 30%-35%, primarily because patients are diagnosed at an advanced stage and have high rates of locoregional and distant recurrence after curative resection[3].

Over the past two decades, the management of locally advanced GC has shifted considerably from surgery-alone to integrated multimodal treatment strategies. Neoadjuvant chemotherapy right before surgical resection has become a cornerstone to this paradigm shift[4,5]. The reasons for neoadjuvant therapy include tumor downstaging to enable complete resection, early intervention of micrometastatic disease, better tolerability than adjuvant chemotherapy, and provision of in vivo chemosensitivity assessment[6,7]. Landmark trials such as the MAGIC and FNCLCC/FFCD 9703 studies provided evidence for perioperative chemotherapy to become standard of care in Western countries, with substantial improvement observed in overall survival (OS) when compared to surgery alone[8,9].

The best neoadjuvant chemotherapy regimen for GC is still being studied. Combining platinum and fluoropyrimidines (5-fluorouracil or capecitabine)-more commonly referred to as platinum-fluoropyrimidine doublet regimens-have long been considered the standard of care, underpinned by a wealth of clinical research and acceptable toxicity[10-12]. Objective response rates (ORRs) for these doublet regimens are conventionally approximately 35%-45%, and are associated with a survival benefit in various phase III studies[13,14]. Nevertheless, the relatively low response rates have pushed for further investigation of more aggressive triplet regimens combining taxanes (paclitaxel or docetaxel) with platinum-fluoropyrimidine scaffolds[15-17].

Taxanes have different modes of action, such as stabilization of microtubules and G2/M cell cycle arrest, so that the combination with platinums or fluoropyrimidines can counteract in cancer antagonism[18,19]. Synergistic interactions between these drug classes have been shown in preclinical studies, with taxane use achieving an augmented platinum DNA adduct burden and countering fluoropyrimidine resistance mechanisms[20,21]. Phase II trials of triplet regimens in the early stages reported favorable response rates above 50%-60% which intrigues their use in the neoadjuvant setting[22-24]. On the other side greater toxicity, especially myelosuppression and peripheral neuropathy have limited their uptake in standard clinical practice[25,26].

Over the last few years, several randomized controlled trials (RCTs) have been published directly comparing triplet and doublet regimens in the neoadjuvant setting for GC with conflicting results. While some studies showed better response rates and potential survival advantage with triplet regimens[27,28], others found no added benefit and adverse treatment-related toxicity was raised as a concern[29,30]. Heterogeneity in trial results may be due to differences in patients, disease characteristics, dosing schedules of the drug, and duration of treatment. In addition, the association between improved pathological response and survival outcomes over the long term is more poorly characterized for GC compared to other solid tumors in which complete pathological response strongly predicts survival[31,32].

In the neoadjuvant setting, where treatment is intended with curative intent, this debate about whether enhanced triplet chemotherapy should justify possible increased toxicity remains relevant and important in terms of preserving quality of life. The trials to date have been individually inadequately powered for survival differences, and a meta-analytic synthesis is thus required to provide definitive evidence. Previous meta-analyses have mainly focused on palliative settings or have pooled preoperative and palliative treatment populations, limiting their applicability to the neoadjuvant setting[33,34]. In addition, reliable safety analyses of particular adverse events (AEs) between regimens have been limited.

Thus, we performed this pooled analysis to comprehensively assess the benefit and safety of taxane-platinum-fluoropyrimidine triplet chemotherapy compared with platinum-fluoropyrimidine doublet chemotherapy in neoadjuvant setting for GC. The primary objectives were: (1) To evaluate the efficacy in terms of ORR and disease control rates (DCRs) of triplet regimens; (2) To assess OS and progression-free survival (PFS) impact; (3) To fully compare safety profiles including grade 3-4 AEs and finally; and (4) To provide evidence-based recommendations for selection of neoadjuvant regimens for clinical practice.

MATERIALS AND METHODS

Search strategy

A systematic literature search was conducted according to the PRISMA guidelines. We searched PubMed (MEDLINE), EMBASE, and the Cochrane Central Register of Controlled Trials electronic databases from their inception to December 31, 2025. Search strategy used a combination of medical subject headings (MeSH) terms, free-text for GC, neoadjuvant chemotherapy and drug class. Main search terms: ("Stomach Neoplasms" OR "Gastric Cancer" OR "Gastric Carcinoma") AND ("Neoadjuvant Therapy" OR "Preoperative Chemotherapy" OR "Primary Chemotherapy") AND ("Taxanes" OR "Paclitaxel" OR "Docetaxel") AND ("Platinum Compounds"[Mesh] or"Cisplatin"[Mesh]or "Oxaliplatin"[Mesh]) and("Fluoropyrimidines"[Mesh] or 5-Fluoro-uracil[mesh]"AND""Capecitabine"[MeSH Terms]) and Randomized Controlled Trial[ptype]. We applied Boolean operators (AND, OR, NOT) to combine search terms. No language restrictions were imposed. Manual screening of the reference lists in studies included and relevant review articles.

Inclusion and exclusion criteria

Inclusion criteria: (1) Study design: RCTs comparing treatment regimens; (2) Population: Patients with histologically confirmed gastric or gastroesophageal junction adenocarcinoma without prior surgical resection, radiotherapy, or systemic chemotherapy; there were no restrictions based on age, sex, race and nationality; (3) Intervention: Compared neoadjuvant chemotherapy taxane-platinum-fluoropyrimidine triplet regimens vs platinum-fluoropyrimidine doublet regimens in experiment arm vs control group arms; and (4) Outcomes reporting at least one of these ORR, DCR, OS, PFS or grade 3-4 AEs per common terminology criteria for AEs.

Tumor response was assessed and defined according to Response Evaluation Criteria in Solid Tumors as complete response (CR), partial response (PR), stable disease (SD), or progressive disease. ORR was defined as the percentage of patients who achieved either a CR or PR, and DCR that consisted of CR, PR, or SD.

Exclusion criteria: (1) Non-randomized studies, including cohort studies, case-control studies, case series, and case reports; (2) Studies exploring neoadjuvant chemotherapy in combination with radiotherapy, targeted therapy or immunotherapy; (3) Studies without full original data or enough information for data extraction; (4) Duplicated publications or pooled patient cohorts; or (5) Review articles, editorials/commentaries, conference abstracts not published as full text articles and study protocols without results.

Study selection and data extraction

Endnote X9 reference management software was used by two independent reviewers to screen titles and abstracts of all retrieved citations. Full text review of potentially eligible studies occurred for final inclusion determination. Disputes were resolved by discussion between a third senior reviewer. Two reviewers independently extracted data using standardized forms. Data were extracted from each study on the following: (1) Study characteristics: First author, publication date, country of origin, type of study design including sample size and inclusion criteria; (2) Patient demographics: Age, male:female ratio, disease stage and performance status; (3) Intervention: Specific chemotherapy regimens used in each group, drug dosing information alongside administration schedule details and number of cycles delivered and completion rates; (4) Outcome measures: ORR, DCR, OS, PFS and hazard ratios (HRs) with corresponding 95% confidence intervals (95%CIs) for OS/PFS data where available alongside frequency of certain grade 3-4 AEs within studies; and (5) Follow up duration along with documented survival data. Kaplan-Meier curves reporting survival outcomes were manually digitised (with Engauge Digitizer 12.1 Software) to extract time-to-event data directly from published survival curves.

Quality assessment

Two reviewers used the Cochrane Collaboration’s Risk of Bias tool to independently assess the methodological quality of included RCTs. The following seven domains were assessed: (1) Random sequence generation (selection bias); (2) Allocation concealment (selection bias); (3) Blinding of participants and personnel (performance bias); (4) Blinding of outcome assessment (detection bias); and (5) Incomplete outcome data, basically how many patients drop out of the study later on in terms of attrition bias; we want a small number 7 selectivity reporting their core points, so this would have been flagged as reporting biases; 8 other potential sources of biases. We assessed each domain as low, high or unclear risk of bias. High quality was defined as low risk in all, and low quality as high risk in one or more key domains with all other studies classified as moderate quality. Quality assessment disagreements were resolved by consensus discussion.

Statistical analysis

RevMan version 5.4 software for Cochrane Collaboration was used to perform the meta-analysis. Statistical heterogeneity among studies was evaluated with the aid of a Cochran Q test (χ² test) and quantified by the described I² statistic (3 or 4). Substantial heterogeneity was defined as P < 0.10 for the Q test or I² ≥ 50%. When heterogeneity was not significant (P ≥ 0.10 and I² < 50%), pooled analysis using the Mantle-Haenszel method with a fixed-effects model was performed. In the presence of considerable heterogeneity, we used a random-effects model by applying the DerSimonian-Laird method with within-study and between-study variance.

Dichotomous outcomes (ORR, DCR and AEs) were quantified in terms of treatment effects as risk ratio (RR), with 95%CIs. HR with 95%CI were calculated for time-to-event outcome measures (OS and PFS). Statistical significance was defined as a two-sided P value < 0.05. The forest plots presented effect sizes and confidence intervals for individual studies and pooled estimates.

Sensitivity analysis was done by leave-one-out method, in that individual study was omitted one at a time to determine the quality and stability of combined results. Results were rated unstable if removal of any study substantially affected the overall estimate or changed statistical significance. Funnel plots were used to evaluate publication bias for outcomes with 10 or more studies, and asymmetry was visually evaluated and assessed by Egger’s regression test when applicable. Statistical significance was defined by two-sided P < 0.05 for all statistical tests.

RESULTS

Literature search and study selection

The systematic search yielded 252 potentially relevant citations. A total of 12 RCTs[4,6,10,19-25,35,36] satisfying inclusion criteria were included in the meta-analysis after deletion of duplicates and screening (Figure 1).

Figure 1  PRISMA flow diagram illustrating the study selection process for inclusion in the meta-analysis.

Study characteristics

The RCTs included[4,6,10,19-25,35,36] published between 2017 and the end of December 31, 2025 involved a total of 11126 patients. They were all phase III trials, and stage II-III disease predominated. Table 1 summarizes the characteristics of the studies.

Table 1 Characteristics of included studies.

Ref.	Sample size (T/D)	Stage	Outcomes
Kang et al[4], 2021	353/466	II/III	ORR, DCR, OS, PFS, AEs
Al-Batran et al[6], 2019	143/205	I/II/III/IV	ORR, DCR, OS, AEs
Zhang et al[10], 2021	76/76	II/III	ORR, DCR, OS, PFS, AEs
Kang et al[19], 2024	749/321	II/III	ORR, DCR, OS, AEs
Al-Batran et al[20], 2016	451/73	I-IV	ORR, DCR, OS, PFS, AEs
Kurokawa et al[21], 2023	60/41	II/III	ORR, DCR, OS, AEs
Park et al[22], 2013	218/232	II/III	ORR, DCR, OS, PFS, AEs
Grizzi et al[23], 2022	107/107	II/III	ORR, DCR, OS, PFS, AEs
Tsuburaya et al[24], 2014	63/262	II/III	ORR, DCR, OS, PFS, AEs
Ni et al[25], 2021	20/45	II/III	ORR, DCR, OS, PFS, AEs
Yoshida et al[35], 2019	109/70	II/III	ORR, DCR, OS, PFS, AEs
Van Cutsem et al[36], 2006	117/92	I-IV	ORR, DCR, OS, PFS, AEs

T: Triplet chemotherapy regimen; D: Doublet chemotherapy regimen; ORR: Objective response rate; DCR: Disease control rate; OS: Overall survival; PFS: Progression-free survival; AEs: Adverse events.

ORR

ORR data were available for 10195 patients from 10 studies. There was significant heterogeneity observed (I² = 84%, P < 0.0001). Triplet therapy showed noticeably higher ORR vs doublet therapy (RR = 1.73, 95%CI: 1.64-1.82, P < 0.0001) indicating a relative advantage of objective response by 73%, as shown in the forest plot at Figure 2A.

Figure 2 Forest plot of objective response rate, disease control rate, overall survival, progression-free survival and grade 3-4 adverse events comparing triplet vs doublet neoadjuvant chemotherapy. A: Objective response rate (ORR). The pooled analysis demonstrated significantly higher ORR with triplet regimens [risk ratio (RR) = 1.73, 95% confidence interval (95%CI): 1.64-1.82, P < 0.0001]. Substantial heterogeneity was present (I² = 84%). Each square represents the point estimate for individual studies, with horizontal lines indicating 95%CIs. The diamond represents the pooled effect estimate; B: Disease control rate (DCR). Triplet regimens achieved significantly superior DCR (RR = 1.22, 95%CI: 1.15-1.28, P < 0.0001) with moderate heterogeneity (I² = 61%). The forest plot displays individual study estimates and the pooled random-effects estimate; C: Overall survival (OS). Despite improved response rates, no significant OS difference was observed [hazard ratio (HR) = 1.24, 95%CI: 0.90-1.72, P = 0.19] with considerable heterogeneity (I² = 96%). HRs less than 1 favor the experimental (triplet) arm; D: Progression-free survival (PFS). No significant PFS benefit was demonstrated for triplet regimens (HR = 1.52, 95%CI: 0.93-2.50, P = 0.09) with substantial heterogeneity (I² = 94%). The wide confidence interval precludes definitive conclusions; E: Grade 3-4 adverse events (AEs). Comprehensive safety analysis demonstrated no significant differences in toxicity profiles between regimens across multiple AEs. The comparable safety suggests good tolerability of triplet therapy with modern supportive care. 95%CI: 95% confidence interval.

DCR

Ten trials with 10195 patients reported a DCR. Heterogeneity was moderate (I² = 61%, P = 0.006). The triplet regimen showed a relative improvement in DCR compared with the doublet chemotherapy regimen: 93.5% vs 87.0%, RR = 1.22, 95%CI: 1.15-1.28, P < 0.0001 (Figure 2B).

OS

OS data were available from nine studies for 9847 patients. Heterogeneity was present (I² = 96%, P < 0.001). Meta-analysis showed no difference of OS between treatment groups (HR = 1.24, 95%CI: 0.90-1.72, P = 0.19). Nonetheless, the broad confidence interval implies both possible benefit and harm of transfusion for survival (as well as equivalence), suggesting that the balance of evidence currently remains unclear rather than proving definitively what is or is not equivalent. This absence of survival benefit remained across all sensitivity analyses (Figure 2C).

PFS

Data for PFS was available from 9 trials (9847 patients). Substantial heterogeneity was observed (I² = 94%, P < 0.001). A random-effects model showed no statistically significant difference in PFS (HR = 1.52, 95%CI: 0.93-2.50, P = 0.09), with a wide confidence interval (Figure 2D).

Safety and AEs

For all evaluated toxicity outcomes, comprehensive analysis of grade 3-4 AEs did not show any statistically significant difference between triplet and doublet regimens. Hematologic toxicities were also similarly frequent: Leukopenia (RR = 1.06, 95%CI: 0.96-1.16, P = 0.26), neutropenia (RR = 1.08, 95%CI: 0.93-1.25, P = 0.32), anemia (RR = 0.98, 95%CI: 0.82-1.17, P = 0.82), and thrombocytopenia (RR = 1.05, 95%CI: 0.87-1.27, P = 0.62). There were also no significant differences in gastrointestinal or constitutional symptoms. The similar safety profile suggests that modern supportive care has ameliorated toxicities of concern from historical perspectives (Figure 2E).

Publication bias

The funnel plot analysis indicated the presence of asymmetry for ORR and DCR (Figure 3), so that it may have a probable publication bias leaving under published small negative studies. Although results were not robust to all sensitivity analyses and inclusion of larger trials, enough findings were consistent that the investigational use is supported. Funnel plots did not reveal extensive bias for survival outcomes without treatment differences.

Figure 3 Funnel plot for objective response rate. Visual asymmetry suggests potential publication bias, with fewer small studies showing null or negative results in the lower portion of the plot. However, the presence of large high-quality trials supports the validity of pooled estimates despite detected bias. OR: Odds ratio.

DISCUSSION

This large-scale meta-analysis of 12 RCTs including more than 11000 patients strongly establishes the relative efficacy and safety end points associated with triplet vs doublet neoadjuvant chemotherapy approaches for GC. Our analysis provides important insights: Taxane-containing triplet regimens improve tumor response rates as well as disease control, but enhanced responses in isolation fail to confer survival benefits and safety is similar between strategies.

The most eye-catching observation was the mighty improvement in ORR (73% relative increase) and DCR (22% relative increase) due to triplet chemotherapy.

These results are clinically significant, as common across individual trials despite geographic and methodologic diversity. In neoadjuvant setting, the increased tumor response provides such possible advantages as: Resectability of borderline tumors, recusal with R0 were negative pathology site margins are predominated along edge tumor size via down upstairs - extent liberation surgery and lower perioperative complications. In patients with bulky or locally advanced tumors in which complete resection may be difficult, the higher response rates associated with triplet therapy may be sufficient to change the operability of a tumor.

Yet, the lack of any survival benefit despite increased response rates is a disturbing disconnect that warrants deliberation. There are different reasons that can justify this paradox. One is that, pathological response in stomach cancer is less likely to be a strong surrogate for long-term survival than breast cancer or rectal cancer. Intensified local-regional therapy may not be sufficient for micrometastatic disease burden, present at diagnosis. Second, the relatively small absolute differences in response rates (usually 10%-15% across individual trials) may not provide meaningful survival benefit against persistent systemic disease biology. Third, potential benefits of survival would be lost because competing events can reduce or dilute the obtainable life-time benefits from treatment (for example postoperative complication, discontinuation of treatment and/or disease progression). Fourth, the follow-up in some of the included trials being relatively short may not have been long enough to identify late-emergent differences in survival.

The comparable safety profiles of triplet vs doublet regimens seen in our meta-analysis challenge the previous dogma that triplet therapy is limited by excessive toxicity. This observation probably mirrors several recent trends: The implementation of patient selection criteria that exclude elderly patients and those with poor performance status; dose modifications and schedule optimization based on pharmacokinetic-pharmacodynamic studies; universal use of granulocyte colony-stimulating factors and other supportive care modalities; and increased recognition and intervention for toxicities associated with taxanes such as hypersensitivity reactions or neuropathy. However, we examined only grade 3-4 toxicities; cumulative lower-grade AEs and quality of life advantages may accrue with doublet regimens, which current data do not adequately address.

Our results have important clinical implications for treatment selection. In patients with clearly resectable tumors, conventional doublet chemotherapy seems to be adequate as it provides a significant clinical benefit along with acceptable tolerability without over-treatment. On the other hand, for patients with borderline resectable or bulky locally advanced disease and where maximal tumor downstaging is critical to achieve complete resection, triplet regimens are a logical consideration due to improved response rates without increasing toxicity burden. Such recommendations are in accord with recent consensus statements from the expert groups, including Japanese Society of Gastroenterological Surgery that highlight tailoring neoadjuvant chemotherapy regimens on tumor features and resectability[37,38]. Many factors such as patient age, comorbidities, performance status, and treatment tolerability should be considered when creating a personalized approach. Additionally, novel biomarkers of chemosensitivity may one day allow for tailored selection of intensified therapy in patients who would respond with avoidance of futile treatment in non-responders.

Compared with previous reports, the global burden of GC continues to evolve, and recent projections to 2050 underscore the need for effective management strategies in this field[39]. Triplet regimens demonstrated symmetric and modest survival benefits in earlier meta-analyses that included patients in the perioperative and palliative settings, but mixtures of treatment intents and patient populations confounded interpretation. The unconditional focus on neoadjuvant therapy in patients with potentially curable disease creates a more relevant evidence base for this specific clinical setting. However, indirect comparisons inevitable carry uncertainty over direct evidence synthesis using our head-to-head evidence synthesis and recent network meta-analyses have been adopted to attempt a ranking of multiple chemotherapy combinations. The changing treatment paradigm for resectable GC highlights the relevance of tailoring regimens according to tumor characteristics and the tumor microenvironment, alongside patient-related factors[40]. The highest level of evidence (individual patient data meta-analysis) and exploration of treatment-covariate interactions is necessary and should collaborative effort is prioritized.

Several limitations warrant acknowledgment. First, significant heterogeneity was observed for most outcomes, likely reflecting differences in patient cohorts, disease characteristics, drug regimens and schedules, supportive care, and surgical methods. Although residual confounding cannot be eliminated, random-effects modeling takes into consideration between-study variance. Publication bias was found for response outcomes, which may overestimate treatment effects, but few large neutral trials were included to alleviate this concern. The third limitation was that included trials had suboptimal reporting of long-term quality of life data and patient-reported outcomes limiting comprehensive benefit-risk assessment. Fourth, molecular heterogeneity in GC has been increasingly recognized and yet subgroup analyses by HER2 status (if positive), microsatellite instability, or Epstein-Barr virus positivity were not conducted here because of lack of data. Fifth, although all centres perform standardized protocols for surgical treatments and those factors which vary from institution to institution (e.g., extent of pelvic and para-aortic lymphadenectomy), it is conceivable that procedures may have inherently affected survival independent of chemotherapy regimen.

Future directions for research glance towards three main areas. This work has two immediate implications: First, it would permit tailored selection of patients for triplet chemotherapy if predictive biomarkers are identified and validated. Examples of candidate biomarkers are tumor mutational signatures, subtypes of the immune microenvironment and pharmacogenomic variants for drug metabolism. Second, novel triplet combinations involving immunotherapy checkpoint inhibitors or targeted agents (anti-HER2, anti-FGFR2) in an effort to increase efficacy and acceptable toxicity. Docetaxel-based triplet regimens have recently been investigated also in some other types of tumors, among them docetaxel + oxaliplatin + S-1[41] for esophagogastric junction adenocarcinoma and locally advanced GC, where the three-drug combination has shown feasibility and encouraging short-term results[41-43]. Third, dedicated trials are required to clarify the optimal sequence and duration of neoadjuvant vs adjuvant chemotherapy. Neoadjuvant use of S-1 plus cisplatin has been investigated in select high-risk populations, such as the JCOG0501 trial[44], whilst better perioperative strategies including chemoradiotherapy have recently been directly compared (ARTIST 2[45]; CRITICS[46]). Fourth, adding circulating tumor DNA dynamics and radiomics-based imaging biomarkers as early indicators of response may allow for the use of adaptive treatment strategies. Fifth, economic evaluations of triplet vs doublet regimens are required for decisions regarding resource allocation impact particularly in low resource settings.

CONCLUSION

While assessed independently in subsequent trials, this meta-analysis shows that taxane-platinum-fluoropyrimidine triplet chemotherapy significantly improves the ORR and DCR over platinum-fluoropyrimidine doublet chemotherapy when used neoadjuvantly in GC. Nevertheless, these improved response rates have not yet been conclusively shown to correspond with OS or PFS benefits, with wide confidence intervals suggesting that clinically relevant effects cannot be excluded and that the evidence remains inconclusive. Most importantly, when comparing the regimens, their safety profiles are similar. These principles validate how we select our treatment: For the great majority of patients with otherwise resectable disease, doublet chemotherapy is appropriate; whereas triplet regimens may be indicated in those who need maximal downstaging before complete extirpative surgery can be contemplated. Further studies combining predictive biomarkers and new drug combinations may optimize neoadjuvant strategies in reshaping survival outcomes of GC.