Intraperitoneal and intravenous paclitaxel plus S-1 and sintilimab as first-line treatment for gastric cancer with peritoneal metastasis: a single-arm phase 2 trial (DRAGON-09)

Summary

Background

First-line PD-1 blockade plus chemotherapy has shown significant clinical benefit in metastatic gastric cancer. This study aimed to evaluate the efficacy and safety of sintilimab and S-1 plus intraperitoneal and intravenous paclitaxel as first-line treatment for patients with gastric cancer peritoneal metastasis (GCPM), as well as exploratory predictive biomarker analysis.

Methods

We conducted a single-arm, phase 2 trial at a single centre in China between Jan 1, 2022, and Dec 31, 2023. Patients with laparoscopically confirmed GCPM were treated with sintilimab (200 mg IV on D1), S-1 (40–60 mg orally twice a day for D1–14), and paclitaxel (20 mg/m² intraperitoneally and 50 mg/m² IV on D1 and 8), in cycles every 3 weeks. The primary efficacy endpoint was overall survival (OS) rate at 1 year. This study is registered with ClinicalTrials.gov, NCT05204173.

Findings

38 patients were included. The median progression-free survival was 14.6 months (95% CI: 10.8–not reached [NR]) and OS was 18.4 months (95% CI: 15.0–NR). The post-hoc analysis showed the objective response rate was 57.9% and disease control rate was 94.7%. The most common treatment-related adverse events were anemia, glutamic oxaloacetic transaminase elevated, and leukopenia. Longitudinal analyses of plasma circulating tumor DNA showed that low baseline human genome equivalent was associated with favorable OS.

Interpretation

First-line sintilimab and S-1 plus intraperitoneal and intravenous paclitaxel showed promising therapeutic efficacy in patients with GCPM.

Funding

Natural Science Foundation of Shanghai, Shanghai Municipal Health Commission, National Science Foundation of China, and the Shanghai Hospital Development Center.

Research in context.

Evidence before this study

We searched PubMed for articles published up to March 31, 2025, with the search terms “gastric cancer”, “peritoneal metastasis”.

Managing peritoneal metastasis in advanced gastric cancer remains a significant clinical challenge, as these patients typically experience rapid disease progression and limited survival. Owing to the presence of plasma-peritoneal barrier, drug delivery from systemic intravenous chemotherapy to peritoneal lesions is often limited, leading to suboptimal treatment outcomes. To address this issue, intraperitoneal chemotherapy has emerged as a promising strategy for local therapeutic intensification.

Although the PHOENIX-GC phase III trial did not formally meet its primary endpoint, its treatment approach-combining intravenous and intraperitoneal paclitaxel with S-1-offered important insights into how therapy might be optimized for patients with peritoneal metastasis.

Globally, phase 3 trials focusing on integrated intraperitoneal and systemic therapy for this population remain limited. China's DRAGON-01 study provided further evidence supporting the value of combined intraperitoneal and systemic chemotherapy in improving outcomes for gastric cancer patients with peritoneal metastasis.

With the growing role of immunotherapy, PD-1 inhibitor-based regimens have become the standard first-line treatment for advanced gastric cancer. The strategy of combining PD-1-based immunotherapy with dual intraperitoneal and intravenous chemotherapy warrants further exploration and validation in prospective clinical studies.

Added value of this study

To the best of our knowledge, this is the first prospective phase 2 trial to explore sintilimab combined with intraperitoneal and intravenous paclitaxel plus S-1 as first-line therapy for patients with gastric cancer peritoneal metastasis. This study showed the promising therapeutic efficacy of sintilimab in combination with systemic therapy in gastric cancer patients with peritoneal metastasis.

Implications of all the available evidence

Our results, together with existing evidence from previous studies, suggest the promising therapeutic efficacy of intraperitoneal and intravenous paclitaxel plus S-1 and sintilimab in patients with gastric cancer peritoneal metastasis. Further validation of the activity from a phase 3 randomized trial is needed to confirm the preliminary results from this phase 2 trial.

Introduction

Globally, gastric cancer ranks as fifth in terms of both incidence and mortality.¹ In China, gastric cancer remains the fifth most common cancer and ranks fourth among causes of cancer-related death.² Peritoneum is the most common site of metastasis of gastric cancer with limited therapeutic options and dismal prognosis. The current standard of care for patients with gastric cancer peritoneal metastasis (GCPM) is systemic chemotherapy ± intraperitoneal chemotherapy ± molecular targeted therapy.³ The PHOENIX-GC study compared intraperitoneal and intravenous paclitaxel plus S-1 versus cisplatin plus S-1 in patients with GCPM and showed that patients with moderate ascites derived benefit from intraperitoneal paclitaxel.⁴ Our previous phase 2 study reported that intraperitoneal and intravenous paclitaxel plus oral S-1 was effective and safe in patients with GCPM.⁵

Targeting the programmed death-1 (PD-1) and programmed death-ligand 1 (PD-L1) interactions has paved the way for a new era of clinical management of patients with gastric cancer. Recently, five phase 3 trials have consistently demonstrated the superior efficacy of PD-1 blockade plus chemotherapy as first-line treatment for patients with advanced or metastatic gastric cancer.6, 7, 8, 9, 10 Furthermore, in RATIONALE-305 trial, the prespecified subgroup analysis of overall survival (OS) in patients with GCPM showed first-line tislelizumab plus chemotherapy could improve the prognosis of these patients, with unstratified hazard ratio of 0.80 (95% CI: 0.65–0.98).⁹ Although the combination of a PD-1 inhibitor and chemotherapy has become a standard of care for unresectable gastric cancer, scarce evidence supports the clinical efficacy of the treatment modality for individuals with peritoneal metastasis.

In addition, advances in tumor biology and next generation sequencing technology enabled the selection of responders to chemoimmunotherapy. However, predictive power of the existing molecular biomarkers, such as PD-L1 expression, microsatellite instability (MSI) status, and tumor mutational burden (TMB) was unsatisfactory.¹¹ The results of SPOTLIGHT and GLOW trial demonstrated Claudin-18 isoform 2 (CLDN18.2) positivity as an emerging molecular biomarker for patients with locally advanced unresectable or metastatic gastric or gastro-esophageal junction (G/GEJ) adenocarcinoma.^12,13 Therefore, stratifying patients who would benefit most from biomarker-directed therapies and developing new approaches to facilitate therapeutic efficiency are highly important.

Here, we conducted this single-arm, phase 2 trial and aimed to evaluate the efficacy and safety of sintilimab and S-1 plus intraperitoneal and intravenous paclitaxel as first-line treatment for patients with GCPM, as well as exploratory predictive biomarker analysis. We found that first-line sintilimab and S-1 plus intraperitoneal and intravenous paclitaxel was promising in patients with GCPM. Longitudinal analyses of plasma circulating tumor DNA (ctDNA) showed that low baseline human genome equivalent was associated with favorable OS.

Methods

Study design and participants

This study (clinicaltrials.gov identifier, NCT05204173) was approved by the Ethics Committee of Ruijin Hospital (Ethics Committee Reference Number 2021 (336)), Shanghai Jiao Tong University School of Medicine, implemented in Ruijin Hospital, China, from Jan 1, 2022 to Dec 31, 2023. Treatment-naïve patients (age ≥18 years at registration) with laparoscopically and histologically confirmed GCPM were eligible, with no history of resection of the primary or metastatic lesions. Other key eligibility criteria included at least one measurable lesion according to Response Evaluation Criteria in Solid Tumors (RECIST) v1.1, Eastern Cooperative Oncology Group (ECOG) performance status ≤1, and adequate organ function. Key exclusion criteria included HER2 positive, distant metastasis other than peritoneal metastasis, and symptomatic central nervous system metastases. The amount of ascites was evaluated by computed tomography and categorized as none, small (within the pelvic cavity), or moderate (beyond the pelvic cavity) at initial enrollment. Peripheral blood samples were collected at baseline, every three cycles of treatment (intraperitoneal and intravenous paclitaxel plus S-1 and sintilimab) concurrent with CT evaluation, and at disease progression sequentially.

Study procedures

Patients received sintilimab (200 mg iv. infusion on day 1), along with intraperitoneal and intravenous paclitaxel plus S-1 treatment (paclitaxel 20 mg/m² intraperitoneally on days 1 and 8, paclitaxel 50 mg/m² intravenously on days 1 and 8, S-1 40–60 mg orally twice a day for days 1–14) every 3 weeks. Intraperitoneal paclitaxel was diluted in 500 ml of normal saline and administered through a port over 1 h, after intraperitoneal administration of 500 ml of normal saline. Treatment response was evaluated every three cycles. The indications for conversion surgery included obvious shrinkage of peritoneal metastasis on second laparoscopy, negative peritoneal cytology, absence of other distant metastases, resectable primary tumor, and good performance status of the patient.^14,15 For patients with a substantial tumor burden and deemed incurable but responding well to treatment, cytoreductive surgery might be feasible after multidisciplinary deliberation. Patients resumed 3–4 weeks postoperatively with protocol-specified treatment. Those with unresectable disease continued the initial regimen. The planned duration of therapy (sintilimab, S-1, intraperitoneal and intravenous paclitaxel) was one year, followed by maintenance therapy (sintilimab, S-1, intraperitoneal paclitaxel) until disease progression, unacceptable toxicity, investigator decision, or patient withdrawal.

Ethics

This study was approved by the Ethics Committee of Ruijin Hospital, Shanghai Jiao Tong University School of Medicine, complied with the Declaration of Helsinki, and registered in clinicaltrials.gov (clinicaltrials.gov identifier, NCT05204173). Ethics Committee Reference Number is 2021 (336). All patients signed the written informed consent prior to enrollment.

Outcome assessments

The primary efficacy endpoint was OS rate at 1 year. Secondary efficacy endpoints included progression-free survival (PFS), OS, and safety. PFS was defined as the time from first dose of study treatment to first progressive disease or death from any cause. OS was defined as the time from first dose of study treatment to death from any cause.

Longitudinal analysis of ctDNA

Concentrations of ctDNA in plasma are usually expressed in haploid genome equivalents per mL (hGE/mL) and might be calculated by multiplying the mean variant allele frequency (VAF) by the input concentration of cfDNA (in pg/mL) and then divided by 3.3 (as one hGE corresponds to 3.3 pg of DNA).¹⁶ Among all cutoffs ranging from 10% to 90%, the baseline plasma hGE cutoff at 40% was found to have the best performance in predicting OS and PFS. Then the baseline plasma hGE was divided into two subgroups, high and low baseline hGE. Detailed methods of DNA extraction and library sequencing, variant calling, and somatic mutation filtering and mutations’ monitoring are described in the Supplementary Methods.

Statistical analysis

The sample size was calculated by PASS software (version 15.0, NCSS, LLC, UT, USA), based on an assumption that the 1-year OS rate will increase from 50% to 75%.^4,6,17,18 A sample size of 38 will be required to provide a power of 90% for a one-sided α at 0.05 significance level, with a dropout rate of 15%. The total accrual duration will be 12 months, with study duration of 24 months.

The Fisher's exact test was used to analyze the independence between the high and low baseline hGE subgroups and the clinical features. Prognostic analysis was primarily conducted using the R packages survival and survminer. Survival curves were estimated using the Kaplan–Meier approach. Hazard ratio (HR) and 95% confidence interval (CI) were estimated using a Cox proportional hazards model, with the p values assessed using the log-rank test. The Wilcoxon signed rank test was employed to compare the hGE value between pairs of patient groups. All statistical analyses were conducted in R version 4.1.3. Two-sided p values < 0.05 were considered statistically significant in all analyses.

Post-hoc analyses were conducted to evaluate objective response rate (ORR) and disease control rate (DCR). Objective response was assessed by both the independent review committee and the investigator. ORR was defined as the proportion of patients achieving complete response (CR) or partial response (PR) according to RECIST v1.1. DCR was defined as the proportion of patients achieving CR, PR, or stable disease (SD).

Role of the funding source

The funder of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report. Min Shi was responsible for the decision to submit the manuscript.

Results

Patient characteristics

Between Jan 2022 and Dec 2023, 41 patients with GCPM were screened. A total of 38 patients were enrolled and received at least one cycle of sintilimab, intraperitoneal and intravenous paclitaxel plus S-1 treatment (Figs. 1 and 2a). Three patients were excluded (two failed to meet the eligibility criteria, one withdrew consent) (Fig. 1). The baseline characteristics were summarized in Table 1. Briefly, seven (18.4%) patients were 60 years or older, and 15 (39.5%) patients were male. Peritoneal cancer index (PCI) levels of 17 patients (44.7%) were 1–10 score, six patients (15.8%) were 11–20 score, eight patients (21.1%) were 21–30 score, and seven patients (18.4%) were 31–39 score at the first laparoscopic exploration. As of the data cutoff date (June 30, 2024), 12 patients were still receiving study treatment. The remaining 26 patients discontinued the trial intervention for the following reasons: loss to follow-up (n = 1), unacceptable toxicities (n = 2), and death (n = 23) (Fig. 1).

Fig. 1.

Trial profile.

Fig. 2.

Study design. a Swimming plot depicted clinical course of 38 patients. Blue triangles indicated partial response, yellow triangles indicated stable disease, red triangles indicated progressive disease, as defined by Response Evaluation Criteria in Solid Tumors (RECIST) v1.1. Survival time was depicted as gray hotizontal lines. The time of surgery was marked with a vertical light blue line, the progression-free survival (PFS) time was marked with a vertical blue line, and the time of death was marked with a vertical black line. Patients 1, 2, 3, 5, 6, 7, 9, 10, 13, 14, 15, 17, 18, 19, 22, 24, 26, 31, 32, 33, 34, 35, and 38 died at the time of the data cutoff. Left panel showed patient ID. Bottom annotation showed duration of time (months). Response to treatment and surgery/non-surgery were also shown. The patients were divided into two groups based on circulating tumor DNA (ctDNA) status observed during the disease course. A circle colored in white indicated that ctDNA was not detected, while a circle in red signified that ctDNA was detected. b and c Kaplan–Meier curves showed PFS and overall survival (OS) of gastric cancer patients with peritoneal metastasis who received sintilimab combined intraperitoneal and intravenous paclitaxel plus S-1 therapy. PFS and OS were measured from first dose of study treatment to first disease progression/death and death, respectively. ctDNA, circulating tumor DNA; PFS, progression-free survival; PR, partial response; SD, stable disease; PD, progressive disease; hGE, haploid genome equivalents.

Table 1.

Baseline characteristics (N = 38).

Patients	No. (%)
Age at diagnosis (years)
<60	31 (81.6%)
≥60	7 (18.4%)
Sex
Male	15 (39.5%)
Female	23 (60.5%)
PCI
1–10	17 (44.7%)
11–20	6 (15.8%)
21–30	8 (21.1%)
31–39	7 (18.4%)
Peritoneal cytology
CY0	12 (31.6%)
CY1	15 (39.5%)
NA	11 (28.9%)
PD-L1 CPS Score
<1	21 (55.3%)
≥1	14 (36.8%)
≥5	9 (23.7%)
≥10	3 (7.9%)
NA	3 (7.9%)
MMR
pMMR	28 (73.7%)
dMMR	0 (0.0%)
NA	10 (26.3%)
MSI
MSS	25 (65.8%)
MSI-H	1 (2.6%)
NA	12 (31.6%)

Clinical outcomes

With the median follow-up of 17.0 months, this study met its primary endpoint, with 1-year OS rate of 76% (95% CI: 63%–91%). The 12-month PFS rate was 54.0% (95% CI: 39%–75%), the median PFS was 14.6 months (95% CI: 10.8–not reached [NR]), and the median OS reached 18.4 months (95% CI: 15.0–NR) (Fig. 2b and c). 22 (57.9%) patients had confirmed PR, 14 (36.8%) had confirmed SD, and 2 (5.3%) had progressive disease (PD) after first-line therapy. In post-hoc analysis, the ORR was 57.9% and DCR was 94.7%. Based on radiological assessment, multidisciplinary team consultation, and personal preference, 20 of the 38 patients proceeded with conversion surgery (18 total gastrectomy and two distal gastrectomy), resulting in a conversion rate of 52.6%. Pathological characteristics in surgery set (n = 20) were summarized in Table 2. Histopathological assessment showed that 85% (17/20) of tumors were poorly differentiated (G3). According to Lauren's classification, 12 patients had diffused type gastric cancer, while others had either intestinal (3/20) or mixed (1/20) type. The numbers of patients who achieved TRG 1, TRG 2, and TRG 3 were 2 (10%), 11 (55%), and 6 (30%), respectively (Table 2).

Table 2.

Pathological outcomes in the surgery set (N = 20).

Patients	No. (%)
Lauren classification
Diffuse type	12 (60%)
Intestinal type	3 (15%)
Mixed type	1 (5%)
NA	4 (20%)
Borrmann
I	0
II	2 (10%)
III	9 (45%)
IV	8 (40%)
NA	1 (5%)
Histologic grade
G1	0
G2-3	3 (15%)
G3	17 (85%)
ypT stage
ypT0	0
ypT1	0
ypT2	3 (15%)
ypT3	9 (45%)
ypT4a	6 (30%)
ypT4b	2 (10%)
ypN stage
ypN0	5 (25%)
ypN1	1 (5%)
ypN2	2 (10%)
ypN3	12 (60%)
TRG
0	0
1	2 (10%)
2	11 (55%)
3	6 (30%)
NA	1 (5%)

Safety profiles

The safety profile was listed in Table 3. Among the 38 treated patients, 36 patients (94.7%) experienced at least one treatment-related adverse events (TRAEs) (Table 3). No grade 4/5 AEs were reported. The most common TRAEs were anemia (29, 76.3%), glutamic oxaloacetic transaminase elevated (28, 73.7%), leukopenia (27, 71.1%), neutropenia (26, 68.4%), and alanine aminotransferase elevated (22, 57.9%). Patients also had potentially immune-related AEs (irAEs) associated with sintilimab: the most common irAEs were immune dermatitis (13, 34.2%, grade 1–3), pruritus (10, 26.3%, grade 1–3), and urinary tract infection (8, 21.1%, grade 1–3). Grade 2–3 immune-related AEs improved after the treatment with thyroxine or corticosteroids. No grade 4 irAEs were observed, and no treatment-related deaths occurred. Two patients discontinued trial intervention (sintilimab) before disease progression due to irAEs: grade 3 type 1 diabetes.

Table 3.

Adverse events.

Grade 1–3	No. (%)
Leukopenia	27 (71.1%)
Neutropenia	26 (68.4%)
Anemia	29 (76.3%)
Thrombocytopenia	7 (18.4%)
Alanine aminotransferase elevated	22 (57.9%)
Glutamic oxaloacetic transaminase elevated	28 (73.7%)
Hypothyroidism	4 (10.5%)
Hyperthyroidism	1 (2.6%)
Type 1 diabetes	2 (5.3%)
Urinary tract infection	8 (21.1%)
Immune dermatitis	13 (34.2%)
Pruritus	10 (26.3%)
Creatinine	7 (18.4%)
Diarrhea	2 (5.3%)

Biomarker analysis

Five variables at baseline were analyzed, including ctDNA content fraction (CCF), copy number instability (CNI), mutant allele tumor heterogeneity (MATH), TMB, and hGE. We further explored the association between mutational parameters and clinical response/conversion surgery. We found that the level of CCF (p = 0.88; Supplementary Figure S1a), CNI (p = 0.57; Supplementary Figure S1b), MATH (p = 0.65; Supplementary Figure S1c), TMB (p = 0.24; Supplementary Figure S1d) could not differentiate responder from non-responder group. There were no differences in pretreatment CCF (p = 0.71; Supplementary Figure S2a), CNI (p = 0.9; Supplementary Figure S2b), MATH (p = 0.88; Supplementary Figure S2c), TMB (p = 0.21; Supplementary Figure S2d) levels between conversion surgery and non-surgery groups.

To assess the prognostic significance of baseline hGE, 38 patients were divided into two subgroups (hGE-high versus hGE-low) according to the median value of hGE at initiation (Fig. 2a). Responders were defined as PR group, while non-responders were defined as SD/PD group. Responders were mainly in hGE-low subgroup, while non-responder cases were mainly in hGE-high subgroup (Fig. 3a). Surgery patients were mainly in hGE-low subgroup, while non-surgery cases were mainly in hGE-high subgroup (Fig. 3b). The median OS in patients with hGE-low versus hGE-high was 27 and 15 months, respectively (HR = 0.31, 95% CI: 0.12–0.79, p = 0.01; Fig. 3c). Patients with conversion surgery had significantly longer OS than those without conversion surgery (median OS: NR versus 13.5 months, HR = 0.13, 95% CI: 0.05–0.36, p < 0.001; Fig. 3d).

Fig. 3.

Prognostic value of haploid genome equivalents (hGE). a The distribution of patients with Responder/Non-responder in baseline (BL)_hGE_high and BL_hGE_low subgroups. b The distribution of patients with conversion surgery/without conversion surgery in BL_hGE_high and BL_hGE_low subgroups. a and b Comparison of independence between baseline hGE_high/baseline hGE_low subgroups and clinical features was performed using Fisher's exact test. c and d Survival curves were estimated using the Kaplan–Meier approach. Hazard ratios (HR) and 95% confidence interval (CI) were estimated using a Cox proportional hazards model, with the p values assessed using the log-rank test. hGE, haploid genome equivalents.

The median PFS in patients with hGE-low_surgery or hGE-high_surgery or hGE-low_non-surgery or hGE-high_non-surgery was 29.5, 23.2, 10.3 and 7.0 months, respectively (Fig. 4a). The median OS in patients with hGE-low_surgery or hGE-high_surgery or hGE-low_non-surgery or hGE-high_non-surgery was NR, 24.4, 20.2 and 12.5 months, respectively (Fig. 4b). The trend of hGE changes showed significant downregulation between baseline blood samples and those taken at 2–3 months (Wilcoxon signed rank test, n = 20 pairs, p = 0.0037, Fig. 4c). The trend of hGE changes also indicated significant downregulation between baseline blood samples and those taken at 4–6 months (Wilcoxon signed rank test, n = 26 pairs, p = 0.00059, Fig. 4d).

Fig. 4.

Exploratory analysis. a Association between baseline (BL)_ haploid genome equivalents (hGE)/surgery status and progression-free survival (PFS). Patients with low-BL_hGE/surgery had significantly longer PFS than those with high-BL_hGE/surgery. b Association between BL_hGE/surgery status and overall survival (OS). Patients with low-BL_hGE/surgery had distinctly longer OS than those with high-BL_hGE/surgery. c Relationship between therapeutic response (responder versus non-responder) and circulating tumor DNA (ctDNA) changes from baseline to 2–3 months. d Relationship between therapeutic response (responder versus non-responder) and ctDNA changes from baseline to 4–6 months. c and d The x-axis represented the number of samples from baseline blood and the 2–3 month or 4–6 month time points, while the y-axis represented the log-transformed hGE value. The middle line in the box was the median, the bottom and top of the box were the first and third quartiles, and the whiskers extended to 1.5 × interquartile range of the lower and the upper quartiles, respectively. The Wilcoxon signed rank test was employed to compare the hGE value between pairs of patient groups. PFS, progression-free survival; OS, overall survival; BL, baseline; hGE, haploid genome equivalents.

To identify the predictive biomarkers for this treatment, we compared responders (n = 22) and non-responders (n = 16) based on their clinicopathological variables. Subsequently, we performed a univariate Cox regression analysis of factors including clinicopathologic variables, key ctDNA features. As shown in Fig. 5, baseline-hGE, conversion surgery, and PCI at first laparoscopy were associated with PFS and OS (Fig. 5a and c). Furthermore, we selected variables with a p-value < 0.1 in the univariate analysis to be included in the multivariate Cox regression analysis. As illustrated in Fig. 5, after adjusting for those confounding factors, baseline-hGE and conversion surgery were independent factors associated with PFS and OS (Fig. 5b and d). Specifically, patients with low baseline-hGE showed greater PFS benefit (HR = 0.22, 95% CI: 0.06–0.76, p = 0.017), with a similar trend observed for the PFS difference in the conversion surgery versus non-surgery arms (HR = 0.08, 95% CI: 0.02–0.46, p = 0.004; Fig. 5b). OS benefit was observed in patients with low baseline-hGE (HR = 0.28, 95% CI: 0.09–0.81, p = 0.019), with a similar trend observed for the OS difference in the conversion surgery versus non-surgery subgroups (HR = 0.07, 95% CI: 0.01–0.29, p < 0.001; Fig. 5d). An exploratory analysis of PFS and OS stratified by PD-L1 CPS score showed a trend of PFS benefit in CPS ≥1, ≥5 population (Supplementary Figure S3).

Fig. 5.

Univariate and multivariate analyses of clinical parameters on progression-free survival (PFS)/overall survival (OS). a and c A univariate forest plot of the relationship between baseline haploid genome equivalents (hGE) and clinical parameters in predicting PFS/OS. b and d The relative contribution of all clinical parameters and baseline hGE to PFS/OS prediction in the multivariate Cox analysis, with hazard ratios (HR), 95% confidence intervals (CI), and p-values indicated. Horizontal lines represented the 95% confidence interval of HR. HRs for OS and PFS were calculated using univariate or multivariate Cox proportional hazards models. The prognostic analysis was conducted using the survival and survminer R packages. PFS, progression-free survival; OS, overall survival; BL, baseline; hGE, haploid genome equivalents; PCI, peritoneal cancer index; TRG, tumor regression grade; PD-L1, programmed death-ligand 1.

Concordance between longitudinal ctDNA profiling and measurable tumor burden

The serial changes of genomic features in ctDNA and tumor volumes are described in two representative cases (Fig. 6). The values of hGE (Fig. 6a and d) and ctDNA VAF (Fig. 6b and e) all decreased after the initial response to the first-line therapy and were kept at low level/not detected during disease control. After disease progression, the parameters continued to increase during the subsequent treatment. We also displayed the CT imaging of these patients at two time points (baseline and PD) (Fig. 6c and f).

Fig. 6.

Representative cases of circulating tumor DNA (ctDNA) dynamics in the comprehensive management of patients with gastric cancer peritoneal metastasis. a and d A circle colored in white indicates that ctDNA was not detected, while a circle in red signifies that ctDNA was detected. The time of surgery is marked with a vertical light blue line, the time of disease progression is marked with a vertical blue line, and the time of death is marked with a vertical black line. b and e The ctDNA mutations were detected at various time points. The maximum variant allele frequency (VAF) for each time point is displayed at the top. c and f Dynamic changes of tumor volumes in two typical cases. ctDNA, circulating tumor DNA; hGE, haploid genome equivalents; VAF, variant allele frequency; BL, baseline; PD, progressive disease.

Discussion

GCPM is one of the most difficult dilemmas in clinical management of gastric cancer. Patients with GCPM had limited therapeutic options and rapid progression after failure of standard of care, resulting in the dismal prognosis. Recently, first-line PD-1 blockade plus chemotherapy has demonstrated the superior efficacy for patients with advanced or metastatic gastric cancer. Our previous study found that intraperitoneal and intravenous paclitaxel plus oral S-1 was effective and safe in patients with GCPM.⁵ Herein, we performed this phase 2 study to evaluate the efficacy and safety of sintilimab and S-1 plus intraperitoneal and intravenous paclitaxel as first-line treatment for patients with GCPM. Our results revealed that first-line sintilimab and S-1 plus intraperitoneal and intravenous paclitaxel showed promising efficacy in patients with GCPM. Biomarker analysis showed that low baseline hGE was associated with favorable OS.

Immune checkpoint inhibitors such as PD-1 blockade have changed the systemic regimen of advanced G/GEJ cancer in recent years. Results from several previous phase 3 studies showed that PD-1 blockade in combination with chemotherapy provided significant improvement in survival over chemotherapy alone in patients with advanced gastric cancer. Meanwhile, results from studies with PD-1 blockade plus chemotherapy showed inconsistent OS benefits in the first-line setting. In CheckMate 649 trial, nivolumab plus chemotherapy showed superior OS benefit versus chemotherapy alone in patients with PD-L1 combined positive score (CPS) of five or more (HR 0.71; p < 0.0001) and in all randomly assigned patients (HR 0.80; p = 0.0002).⁶ In ATTRACTION-4 trial, nivolumab combined with oxaliplatin-based chemotherapy was not shown OS benefit versus placebo plus chemotherapy in Asian patients (HR 0.90; 95% CI: 0.75–1.08; p = 0.26).¹⁹ In this trial, sintilimab, intraperitoneal and intravenous paclitaxel, plus S-1 regimen showed promising clinical activity and manageable safety profile in treatment-naïve patients with GCPM. The variability in OS among these studies could potentially be attributed to differences in study design, the proportion of Asian compared with European/North American participants, choices of chemotherapy regimens, differences in subsequent therapies, along with the statistical considerations.

To date, there is very limited effective treatment options for patients with GCPM. Given the promising efficacy and mild toxicities, immune checkpoint inhibitor (ICI)-based combination therapy may have a good application prospect. In RATIONALE-305 study, prespecified subgroup analysis of OS results favored patients in tislelizumab plus chemotherapy arm versus placebo plus chemotherapy arm in patients with GCPM.⁹ In the present study, we found that sintilimab plus chemotherapy was effective with the estimated OS rates at 12-month and 24-month of 76% (95% CI: 63–91%) and 38% (95% CI: 25–60%). Collectively, these findings suggest that first-line PD-1 blockade combined with chemotherapy should be considered as the standard of care for patients with GCPM.

In comparison to previous studies of chemotherapy plus PD-1 inhibitors, the toxicity profile of sintilimab combined with chemotherapy regimen in this trial was generally similar. Among the TRAEs, anemia, glutamic oxaloacetic transaminase increased, leukopenia, neutropenia, and alanine aminotransferase increased warrant further attention. It should be noted that two patients of newly diagnosed type 1 diabetes were observed in our study, but were manageable by supportive treatment. The difference in safety profile might be partly explained by different tumor types and ICIs used. However, considering that the evaluation of symptomatic adverse events is characterized by a certain level of subjectivity, further comparative trials are needed.

Further investigations are needed to evaluate the impact of paclitaxel on immune cells and tumor cells. Yu P et al. illustrated genomic and immune microenvironment features influencing chemoimmunotherapy (sintilimab, paclitaxel, and S-1) response in gastric cancer with peritoneal metastasis.²⁰ Li Y et al. reported CAF-macrophage crosstalk in tumor microenvironments governed response to immune checkpoint blockade in gastric cancer peritoneal metastases.²¹ In addition, different chemotherapeutic agents might lead to distinct properties of tumor microenvironment that may influence responses to accompanying ICIs. It is worth to illustrate the mechanistic distinctions of paclitaxel or oxaplatin when combined with sintilimab for metastatic gastric cancer treatment.

Our previous study reported that ctDNA could be used as a potential predictor for response to chemotherapy and survival outcomes in patients with GCPM.²² Therefore, we explored the predictive/prognostic value of longitudinal ctDNA analysis in predicting clinical outcomes of this study regimen. We found that low baseline hGE was independently correlated with better clinical outcomes, although this observation should be viewed as exploratory the modest sample size. Recently, a prospective observational study (NCT02887612) evaluated the predictive value of ctDNA for disease recurrence in locoregional gastric cancer.²³ Residual ctDNA after adjuvant chemotherapy effectively predicts high recurrence risk in patients with stage II-III gastric cancer, and the combination of tissue-based and circulating tumor features could better predict outcomes.²³ Collectively, these findings suggest that ctDNA might be a potential biomarker to predict progression, which needs to be further verified.

There are several limitations of the study. Firstly, the limited sample size may introduce the potential selection bias, thereby resulting in low overall quality of evidence. Secondly, although some patients showed survival benefit from PD-1 blockade combined with chemotherapy, the lack of mechanistic analyses hindered the identification of prognostic indicators for responders/non-responders to the combination therapy. Finally, the current study provides immature data for safety and efficacy, death events reported in 60.5% (23/38) of all enrolled patients at data cutoff. Further validation of the activity from a phase 3 randomized trial is needed due to the preliminary results from this phase 2 trial.

In conclusion, the present study demonstrated that first-line sintilimab and S-1 plus intraperitoneal and intravenous paclitaxel could significantly improve PFS and OS in patients with previously untreated HER2-negative GCPM, with a manageable safety profile. This regimen could be considered as an alternative treatment option for patients with GCPM.

Contributors

Conception and design: Hong Yuan, De-Bin Sun, Zhong-Yin Yang, Min Shi, Zheng-Gang Zhu, Chao Yan.

Provision of study materials or patients: Xue-Xin Yao, Wen-Tao Liu, Ya-Nan Zheng, Zi-Chen Hua, Zhen-Tian Ni, Chang-Yu He, Zhen-Qiang Wang, Chen Li, Jun Zhang, Min Yan.

Collection and assembly of data: Hong Yuan, Sheng Lu, De-Bin Sun, Xue-Xin Yao, Wen-Tao Liu, Ya-Nan Zheng, Zi-Chen Hua, Zhen-Tian Ni, Chang-Yu He, Zhen-Qiang Wang, Zhong-Yin Yang, Min Shi, Zheng-Gang Zhu, Chao Yan.

Data analysis and interpretation: Hong Yuan, De-Bin Sun, Jiao Zhang, Di Liu, Min Shi, Zheng-Gang Zhu.

Manuscript writing: Hong Yuan, Sheng Lu, Jiao Zhang, Min Shi, Zheng-Gang Zhu, Chao Yan.

Final approval of manuscript: All authors.

Accountable for all aspects of the work: All authors.

Chao Yan, Min Shi, and Zhong-Yin Yang had accessed and verified the data.

Data sharing statement

All data supported the findings in this paper are available upon reasonable request via the corresponding author.

Declaration of interests

De-Bin Sun declares employment with Genecast Biotechnology Co. Ltd. Jiao Zhang declares employment with Genecast Biotechnology Co. Ltd. Di Liu declares employment with Genecast Biotechnology Co. Ltd. No other potential conflicts of interest were reported.