The Neo-PLANET phase II trial of neoadjuvant camrelizumab plus concurrent chemoradiotherapy in locally advanced adenocarcinoma of stomach or gastroesophageal junction

Patients’ characteristics and treatment

Between September 14, 2018, and December 22, 2020, 41 patients were screened and 36 patients were enrolled (Fig. 1). Characteristics of 36 patients are listed in Table 1. All patients had lymph node involvement, 30 (83.3%) had T4a disease, and 19 (52.8%) had primary tumors located at GEJ. Twenty-five (69.4%) patients were assessed for PD-L1 expression, and nine had a combined positive score (CPS) of 1 or more. Twenty-seven (75%) patients were assessed for Epstein-Barr Virus (EBV) status, and only one had an EBV-positive tumor. Thirty-three (91.7%) patients were assessed for microsatellite instability (MSI) and tumor mutational burden (TMB) status; all the tumors were MSI-low/microsatellite stability (MSS), and one was TMB-high.

All patients received neoadjuvant therapy and thus were included in the FAS; 32 patients completed neoadjuvant therapy as planned, while one discontinued camrelizumab due to disease progression and three discontinued camrelizumab due to adverse events (AEs): one had neuritis presented with headache and tongue movement disorder, one had myositis and myocarditis with impaired heart function, and one had severe reactive cutaneous capillary endothelial proliferation widely distributed throughout the body. Thirty-three patients underwent total gastrectomy and D2 lymph node dissection, including 30 with open surgery and three with laparoscopic surgery. One patient with open surgery also had a combined splenectomy. Six (18.2%) of 33 patients had delayed surgery: five due to immune-related AEs (irAEs) and one due to non-irAEs (prolonged grade 2 nausea and vomiting). The median operative time was 186 min (range, 140–259). The median intraoperative blood loss was 100 mL (range, 20–400). The median number of lymph nodes resected was 28 (range, 6–51). The median length of postoperative hospital stay was 8 days (range, 6–20). Three patients did not undergo gastrectomy because one had progressive disease, one refused surgery, and one was deemed uncurable due to liver metastases during surgical exploration.

Efficacy

In the FAS, 12 of 36 patients had pathological downstaging to ypT0N0, with both pCR (primary endpoint) and total pCR (tpCR; secondary endpoint) rates of 33.3% (95% CI, 18.6–51.0). In the surgery set, both pCR and tpCR rates were 36.4% (95% CI, 20.4–54.9). The major pathological response (MPR; secondary endpoint) rate was 44.4% (95% CI, 27.9–61.9) in the FAS and 48.5% (95% CI, 30.8–66.5) in the surgery set (Table 2). Four of eight patients with grade 2 tumor regression had just 10% residual tumor cells (Fig. 2a). Representative changes in tumor size by imaging examinations and postoperative pathological images from one patient with pCR (NP025) and another patient with a minor pathological response (NP018; grade 3 tumor regression) are shown in Fig. 2b. Representative pathological features²¹ after immunotherapy are shown in Fig. 2c. The downstaging (secondary endpoint) results showed that 28 (77.8%) of 36 patients reached ypN0 (Supplementary Table 1). Patients with intestinal type had higher pCR rate than those with other types (56.3% [9/16] vs. 17.6% [3/17], P = 0.032). The pCR rates for patients with tumor at GEJ and stomach were 37.5% (6/16) and 35.3% (6/17; P > 0.999; Supplementary Table 2), respectively. Of five patients with delayed surgery due to irAEs, four achieved pCR, and one had grade 3 tumor regression (80% residual tumor cells). The margin-free (R0) resection rate (secondary endpoint) was 91.7% in the FAS and 100% in the surgery set.

By the data cutoff date on February 20, 2022, the median follow-up time was 26.3 months (range, 11.3–41.8). Both median progression-free survival (PFS; secondary endpoint) and overall survival (OS; secondary endpoint) were not reached. The post-hoc analyses showed that the 2-year PFS rate was 66.9% and the 2-year OS rate was 76.1% (Fig. 3b–g). The 2-year PFS and OS rates were 60.0% and 73.1% in patients with tumors at GEJ, and 75.5 and 80.2% in patients with tumors at the stomach (Supplementary Fig. 1), respectively. As demonstrated in Fig. 3a, eight patients who underwent gastrectomy experienced recurrence, and five of them died. Two patients who suffered tumor progression during neoadjuvant therapy died of progressive disease. One patient free from progression died by accident and no recurrence was determined.

Safety

Treatment-emergent AEs (TEAEs) of any grade occurred in all patients during neoadjuvant therapy. Grade ≥3 TEAEs occurred in 28 (77.8%) patients. The most common grade 3 or 4 TEAEs were decreased lymphocyte count (27 [75.0%]) and decreased white blood cell count (two [5.6%]; Table 3). IrAEs occurred in 31 (86.1%) patients, but most were grade 1–2. The most common irAEs were reactive cutaneous capillary endothelial proliferation (25 [69.4%]), thyroid dysfunction (six [16.7%]), hyperglycemia (six [16.7%]), and pruritus (five [13.9%]; Table 4). No AEs leading to death occurred.

Among 33 patients with surgery, surgical complications occurred in 13 (39.4%) patients. The most frequent grade II or III surgical complications were pleural effusion (four [12.1%]), atelectasis (three [9.1%]), and intra-abdominal infection (three [9.1%]; Supplementary Table 3). One patient with intra-abdominal bleeding and one patient with intra-abdominal infection suffered reoperation, both recovered eventually. No grade IV or V surgical complications occurred. No patients had surgical complications leading to readmission within 30 days.

Biomarkers

The pCR rate in PD-L1-positive tumors (defined as CPS at a cutoff point of 1, 5, or 10, respectively) was not significantly higher than in PD-L1-negative tumors (Supplementary Table 4). Analysis of somatic mutation by whole-exome sequencing (WES) of treatment specimens showed a significantly higher pCR rate in patients with pretreatment TMB ≥ median level (4.04 mutation/Mb) than those with TMB < median level. All six patients with intestinal type and TMB ≥ median level achieved pCR (Supplementary Table 4). For MSI and EBV, no tumor was identified as MSI-high by WES and only one tumor was found EBV-positive, so we could not provide information that the MSI-high or EBV-positive was related to tumor response in this study. The only patient with an EBV-positive tumor had grade 3 tumor regression (80% residual tumor cells). The PD-L1 density confirmed by immunohistochemistry (IHC) significantly increased after neoadjuvant therapy (Supplementary Fig. 2).

Study design and participants

In this single-arm phase II clinical trial, patients were enrolled from Zhongshan Hospital, Fudan University between September 14, 2018, and December 22, 2020. Eligible patients were aged 18–75 years; had Eastern Cooperative Oncology Group (ECOG) performance status of 0–1; had histological confirmed adenocarcinoma located at stomach or GEJ (Siewert type II or III); had clinical stage T3-4aN + M0 by endoscopic ultrasound or enhanced computed tomography (CT)/magnetic resonance imaging (MRI) according to the eighth edition of the American Joint Committee on Cancer TNM staging system³⁷; had treatment-naïve disease with the possibility for radical surgery; did not have severe comorbidity that might lead to expected survival less than 5 years, and had adequate organ function. All patients underwent staging laparoscopy to exclude peritoneal metastasis before enrollment. The study protocol listing the full eligibility criteria is provided as Supplementary Note 1 in the Supplementary Information file.

Procedures

Patients received induction chemotherapy with oxaliplatin 130 mg/m² on day 1 and capecitabine 1000 mg/m² twice per day on days 1–14 (XELOX regimen) for a 21-day cycle. Five weeks of concurrent chemoradiotherapy was performed within one week after the completion of induction chemotherapy. Intensity-modulated radiotherapy was performed using a VersaHD accelerator (Elekta, Stockholm). Patients were fixed in a supine position by a vacuum-forming mold and scanned with CT. Images were transferred to the radiotherapy planning system (Monaco, Elekta CMS, Maryland Heights, MO, USA and Pinnacle3, Philips Medical Systems, Cleveland, Inc.). The thickness of each scan layer was 3 mm. Gross tumor volume (GTV) consisting of the primary tumor and visible metastatic lymph nodes was delineated based on CT/MRI and gastroscope. Clinical target volume (CTV) comprised of GTV and elective regional lymph nodes at high risk, and planning target volume (PTV) covered CTV and 5–10 mm beyond its margin. A dose-volume histogram was used to optimize the dose-distribution plan. The total irradiation dose of intensity-modulated radiotherapy was 45 Gy for PTV, delivered in 25 fractions, with 1.8 Gy on days 1–5 each week. Simultaneous capecitabine (850 mg/m² twice per day) was given on each day of radiotherapy. Patients received consolidation chemotherapy with XELOX regimen for a cycle at 2–3 weeks after concurrent chemoradiotherapy and were reevaluated 1–3 weeks later. Camrelizumab 200 mg once every 3 weeks was given five times since the initiation of induction chemotherapy until 3 weeks before surgery. Patients with resectable disease proceeded to gastrectomy and D2 lymph node dissection, the extent of resection was according to Japanese gastric cancer treatment guidelines (version 4)³⁸. Adjuvant chemotherapy with the XELOX regimen was recommended for four cycles at 4–6 weeks after surgery.

Dose reduction for capecitabine and oxaliplatin and delay for the three study drugs and radiotherapy were allowed according to the AEs during the neoadjuvant therapy period. The details of dose adjustment are described in the study protocol, provided in Supplementary Note 1. If disease progression, distant metastasis, or intolerable toxicity occurred before surgery, patients would discontinue the study treatment and switch to other appropriate treatment, with follow-up for PFS and OS.

Radiological assessment was performed by CT/MRI according to the Response Evaluation Criteria In Solid Tumors version 1.1 at baseline, after the completion of concurrent chemoradiotherapy, before surgery, every 6 months after surgery until 3 years, and annually thereafter. The pathological response was assessed according to Becker’s tumor regression grading system³⁹. For the primary tumor, if the macroscopic tumor was identifiable or the area of the stomach with scarring could indicate the site of the previous tumor (tumor bed), the tumor or tumor bed would be cross-sectioned serially at 0.5 cm intervals. If the macroscopic tumor was unidentifiable, the tumor bed would be indicated by baseline gastroscopy and CT, and cross-sectioned serially at 0.5 cm intervals. For lymph nodes, surgeons separated the lymph nodes from the specimen and grouped them into separate stations before formalin-fixation. If the short diameter of a lymph node was more than 1 cm, this lymph node would be divided into two and embedded separately. Other lymph nodes were embedded with a maximum of 4 lymph nodes in one cage. The lymph nodes were cross-sectioned serially at 0.2–0.3 cm intervals. These tissue sections were stained with hematoxylin and eosin (HE) staining, elastic von Gieson staining, and periodic acid–Schiff (PAS) staining. The von Gieson staining was used to distinguish between tumor desmoplasia and scarring as a result of chemotherapy, and the PAS staining was used to help distinguish signet ring cells from histiocytes. IHC for cytokeratins was used when epithelial cells were unidentifiable. If no residual tumor cells were detected, additional three-step sectioning would be performed to confirm pCR.

AEs since the initiation of neoadjuvant therapy until 28 days after the last dose of neoadjuvant therapy were graded according to the National Cancer Institute Common Terminology Criteria for Adverse Events version 4.03. Intraoperative and postoperative surgical complications were graded according to the Clavien-Dindo classification⁴⁰.

Endpoints

The primary endpoint was pCR (ypT0) rate, defined as the proportion of patients with grade 1a (no residual tumor) regression at the primary tumor. Secondary endpoints included tpCR (ypT0N0) rate (defined as the proportion of patients with grade 1a regression at both primary tumor and lymph nodes), MPR rate (defined as the proportion of patients with grade 1a and 1b [<10% residual tumor cells] regression at primary tumor), R0 resection rate, downstaging, PFS (defined as the time from the initiation of neoadjuvant therapy to disease recurrence, progression, or any-cause death), OS (defined as the time from the initiation of neoadjuvant therapy to any-cause death), and safety.

Biomarker analysis

Tumor specimens from pretreatment gastroscopy biopsy and surgical resection were collected for biomarker analysis. The PD-L1 expression level was measured by the 22C3 pharmDx kit (Agilent Technologies, catalog number M3653, clone number 22C3, 1:50 dilution) on the Dako Autostainer Link 48 system (Agilent Technologies). CPS was used to characterize the PD-L1 expression, calculated as the number of all PD-L1-positive cells (tumor cells, lymphocytes, and macrophages) divided by the number of all tumor cells ×100. EBV status of the tumor was determined on available formalin-fixed paraffin-embedded (FFPE) tissue sections by EBV-encoded small RNA in situ hybridization kit (Talent biomedical, catalog number CISH-EBER-100) according to the manufacturer’s instructions.

IHC slides were scanned and digitalized using a Scanscope XT system (Aperio/Leica Technologies). Single-staining IHC quantification analysis was performed by the pathologist using HALO 3.2 software (Indica Labs). The number of marker-positive cells for each analysis area were calculated and expressed as density (number of positive cells/mm²).

TMB and MSI status were determined by WES. Tumor DNA was isolated from FFPE sections of specimens. Maxwell 16 FFPE Plus LEV DNA Purification Kit (Promega, catalog number AS1135) was used to extract FFPE DNA after FFPE sample sections were scalpeled and deparaffinized. Total DNA from tissues (or blood) was extracted using DNeasy 96 Blood & Tissue Kit (Qiagen, catalog number 69504) or Blood Kit (Qiagen, catalog number 51104) according to the manufacturer’s instructions. About 300 ng high-quality genomic DNA concentrations (OD260/280 = 1.8–2.0) were sheared to target of 150–200 bp average size with Covaris LE220 Sonicator (Covaris) and were prepared into DNA libraries using SureselectXT reagent kit (Agilent, catalog number G9611A). A paired-end DNA sequencing library was prepared via end-repair, purifying, A-tailing, paired-end adapter ligation, and amplification. DNA concentration of prepared libraries was measured using a Qubit 3.0 fluorometer (Thermo Fisher Scientific) and the size distribution of the resulting sequencing libraries was analyzed using Agilent Bioanalyzer 4200 (Agilent). Paired-end sequencing is performed using an Illumina NovaSeq 6000 system following Illumina-provided protocols for 2 × 150 paired-end sequencing. Mutations were analyzed using Genome variant caller (GVC) software with default parameters and were annotated using VEP software (Ensembl Variant Effect Predictor, release_100.2)⁴¹. TMB was defined as the number of nonsynonymous variations with an allele frequency of at least 5% in the captured coding region, and the mutational type contained single nucleotide variants and indels. TMB-high was defined as ≥10 mutation/Mb. MSIsensor algorithm⁴², which identifies the percentage of microsatellite loci that are unstable in the tumor genome compared to its matched normal, was used to analyze the MSI status of each sample. MSI status was classified as MSI-high if the proportion of MSI exceeds 10%.

Statistics and reproducibility

This was a single-arm study, and no randomization was used. Biomarker analyses and HE staining of pathological images were performed one time in each independent sample. No data were excluded from the statistical analyses. The investigators were not blinded to allocation during the study period. The pathological response was assessed by independent pathologists who were blinded to patient information.

Simon’s min-max two-stage design was adopted to calculate the sample size. We assumed that the pCR rate with neoadjuvant chemoradiotherapy as historical control was 16%^4,31,43. The addition of camrelizumab to chemoradiotherapy would improve the pCR rate to 35%, with one-sided α of 5% and power of over 80%. In the first stage, if three or more of the 15 patients achieved pCR, another 21 would be accrued to the second stage. If ten or more of the 36 patients achieved pCR, the study treatment would be deemed worthy of future study.

The FAS included all patients who received at least one dose of study treatment. The surgery set included all patients who underwent gastrectomy. Efficacy was analyzed in the FAS and surgery set. AEs during neoadjuvant therapy were analyzed in the safety set, including all patients who received at least one dose of study treatment. Surgical complications were analyzed in the surgery set. Continuous data were expressed as median (range), and categorical data were expressed as frequency (percentage). The 95% CIs of pCR, tpCR, and MPR rates were calculated using the Clopper-Pearson method. PFS and OS were estimated using the Kaplan–Meier method, and post-hoc analyses were performed to calculate 2-year PFS and OS rates. Comparisons of pCR rate in different subgroups were performed using Fisher Exact test, and PFS and OS were compared using a log-rank test. Subgroups were divided by a pathological response (pCR vs. non-pCR; MPR vs. non-MPR), Lauren’s classification (intestinal type vs. diffuse type vs. mixed type vs. unspecified type; intestinal type vs. others), primary tumor site (GEJ vs. stomach), baseline CPS (≥1 vs. <1; ≥5 vs. <5; ≥10 vs. <10), and baseline TMB level (≥median vs. <median). A comparison of PD-L1 density before and after neoadjuvant therapy was performed using paired t-test. Two-sided P < 0.05 was considered statistically significant. Data were collected using Microsoft Office Excel 2010. Data analysis was performed using R software (version 4.0.3).

Peer review information

Nature Communications thanks Russell Petty and the other, anonymous, reviewer(s) for their contribution to the peer review of this work.