2.1 Participants

EXELOX (ClinicalTrials.gov identifier: NCT02395640) was an open-labeled, multicenter, randomized, prospective phase III trial conducted at seven sites in China designed to demonstrate the non-inferiority of the XELOX regimen to the EOX regimen as the first-line treatment for AGC patients. AGC included gastric cancer with primary lesion that cannot be radically resected or metastatic lesions. The protocol, attached in the Supplementary file, was approved by the local institutional review board and ethics committee at each participating site. The trial was conducted in accordance with the International Conference on Harmonization Guidelines for Good Clinical Practice and the Declaration of Helsinki and Chinese law. All patients signed an informed consent form before enrollment in the trial.

Eligible patients had histologically confirmed, unresectable, locally advanced or metastatic adenocarcinoma of the stomach or gastroesophageal junction. Eligible patients had not received chemotherapy for locally advanced or metastatic disease and had at least one measurable lesion according to the Response Evaluation Criteria in Solid Tumors (RECIST 1.1 criteria). Other key inclusion criteria were age of 18 years or older, an Eastern Cooperative Oncology Group (ECOG) performance status of 0-2, adequate organ function, and life expectancy of more than 3 months. Key exclusion criteria included HER2-positive patients who were able to afford and willing to receive trastuzumab treatment, and symptomatic brain or leptomeningeal metastases. Full eligibility criteria are provided in the protocol (Supplementary file).

The intention-to-treat (ITT) set included all patients who were enrolled and randomized, irrespective of whether they received the study medications or not. The per protocol (PP) set included patients who met the eligibility criteria, received at least one cycle of chemotherapy, and underwent at least one tumor response evaluation.

2.2 Randomization and treatment

We randomly assigned eligible patients (1:1) to receive either the XELOX (oxaliplatin 130 mg/m² on day 1 and xeloda 1000 mg/m² on days 1-14) or EOX regimen (epirubicin 50 mg/m² on day 1, oxaliplatin 130 mg/m² on day 1, xeloda 1000 mg/m² on days 1-14). A computer-generated randomization schedule managed by King Yee Company (Beijing, China) was used. We stratified randomization by ECOG status, extent of disease (locally advanced or metastatic), and clinical trial center. Chemotherapy was repeated every 3 weeks until disease progression, intolerable adverse events (AEs), patient's death, withdrawal of informed consent, or up to eight cycles, followed by xeloda single-agent maintenance. The dose modifications are detailed in the trial protocol (Supplementary file).

2.3 Outcomes

The primary endpoint was PFS (the duration from the date of randomization to any documented tumor progression or death due to any cause). The secondary endpoints included OS (the duration from the date of randomization to the date of death due to any cause), ORR (the proportion of patients whose tumor decreased to a certain size and remained for a certain period of time, including those with complete response and partial response), safety, and QoL. Tumors were assessed radiologically at baseline, and efficacy evaluation was conducted every 6 weeks according to the criteria of RECIST 1.1. The global health/QoL (EORTC QLQ-C30) score was assessed at baseline (7 days before the first dose of study drugs) and every 6 weeks. AEs were classified and graded using the Common Terminology Criteria for Adverse Events (version 4.0).

2.4 Statistical analyses

The primary objective was to determine the non-inferiority of the XELOX regimen to the EOX regimen in terms of PFS. Assuming an mPFS of 7.0 months in the EOX arm, referring to the results of the REAL-2 trial [13], a hazard ratio (HR) < 1.3 was considered non-inferior, with 80% power and a one-sided P = 0.05. Accordingly, an estimated 360 progression events were needed to show non-inferiority, and an enrollment of at least 438 patients, accounting for a 10% dropout rate, was set for this trial. In the two previous phase III trials, before the start of the present trial, the non-inferiority margin in PFS of AGC patients under first-line treatment were 1.25 and 1.4 months, respectively [9, 18]. The non-inferiority margin in PFS in the present study was defined in reference to the trial of SOX non-inferior to SP, in which HR < 1.3 was set [18]. Nevertheless, in a phase III trial showing that the modified DCF (mDCF) triplet regimen was superior to the CF doublet regimen in terms of PFS (mPFS: 7.16 vs. 4.93 months) in Chinese patients in the REAL-2 trial [20], in which the mPFS of patients receiving mDCF (7.16 months) was close to that of patients receiving EOX (7.0 months), the HR of PFS for mDCF versus CF was 0.58, and the reciprocal was 1.72, which led to 1.36, with a 50% retention of 1.72. Therefore, the HR margin of 1.30 in the present study was considered suitable for evaluating non-inferiority.

We analyzed PFS, OS, and ORR in the ITT set, and analyzed PFS, OS, and safety in the PP set. OS and PFS were estimated using the Kaplan-Meier method. The log-rank test was used to compare the PFS and OS between the two groups. The differences in ORR and AEs between the groups were compared using the χ² test. The Cox proportional hazards model was used to calculate HRs with 95% confidence intervals (CIs) and P_interaction values for predefined subgroup analyses. QoL scores were calculated according to the procedures defined in the EORTC QLQ-C30 scoring manual. QoL scores were compared using a two-sample t-test or Wilcoxon rank sum test between arms. We also used the last observation carried forward (LOCF) method and Mean-value method to fill in the missing data. The analysis of the primary endpoint PFS was one-sided, and all other statistical analyses were two-sided. The statistical significance level was set at 0.05. Statistical analyses were all performed using the SAS statistical software (version 9.4; SAS Institute Inc, North Carolina State University, USA and Stata statistical software (version 12.0; StataCorp, College Station, TX, USA).

3.1 Patients and treatment

Between April 10, 2015 and August 20, 2020, 448 patients were enrolled (Figure 1) and randomly assigned to the XELOX (n = 222) or EOX arm (n = 226). The baseline characteristics were well balanced between the two arms (Table 1). Two patients in the XELOX arm and 3 in the EOX arm did not receive treatment and were excluded from the PP set. Seven patients in the XELOX arm and 8 in the EOX arm, who were eventually considered ineligible, withdrew due to personal reason, or received only one cycle of treatment without response evaluation, were also excluded from the PP set. One patient was diagnosed with gastric adenocarcinoma at another hospital which did not participate in this trial, but the final histological diagnosis was diffuse large B-cell lymphoma at a participating hospital. For another patient, imaging specialists considered that the lesion of the lung was unmeasurable.

As a result, 428 randomly assigned patients were included in the PP set, with 213 assigned to the XELOX arm and 215 assigned to the EOX arm. The median number of cycles administered to the two arms was six. The rate of dose delay was 32.9% (70/213) in the XELOX arm and 35.3% (76/215) in the EOX arm. Dose modifications resulting from toxicity were more common with EOX than with XELOX (55.8% vs. 35.2%, P = 0.009). The mean relative dose intensity of oxaliplatin (0.866 vs. 0.807) and xeloda (0.877 vs. 0.802) was higher in the XELOX arm than in the EOX arm. The reasons for stopping intravenous chemotherapy are shown in Supplementary Table S1.

For the patients in the PP set who had disease progression after first-line treatment, 47.5% (94/198) in the XELOX arm received second-line treatment and the information of second-line treatment was unavailable for 21.2% (42/198) patients. In the EOX arm, 41.3% (83/201) received second-line treatment and 18.4% (37/201) were lost to follow-up. The paclitaxel-based regimens were most commonly used for second-line treatment. Besides, in the XELOX arm, 59 patients received best support treatment, 2 received resection of primary lesion, and 1 received radiotherapy of primary lesion; in the EOX arm, 68 patients received best support treatment, 5 received resection of primary lesion, 4 received radiotherapy of primary lesion, and 4 rechallenged oxaliplatin and 5-fluorouracil.

3.2 Efficacy

At the time of data cutoff (February 17, 2021), 198 (89.2%) patients in the XELOX arm and 201 (88.9%) in the EOX arm had disease progressed or died. In the ITT set, mPFS was 5.0 months (95% CI = 4.5-6.0 months) in the XELOX arm and 5.5 months (95% CI = 5.0-6.0 months) in the EOX arm (HR = 0.989, 95% CI = 0.812-1.203; P_{non-inferiority} = 0.003; Figure 2A). The upper limit of the 95% CI was lower than the non-inferiority margin of 1.30. The PP set also demonstrated a non-inferiority of PFS in the XELOX arm compared with the EOX arm. The mPFS was 5.0 months (95% CI = 5.0-6.0 months) with XELOX and 5.5 months (95% CI = 5.0-6.0 months) with EOX (HR = 0.983, 95% CI = 0.807-1.198; P_{non-inferiority} = 0.003; Figure 2B).

After a median follow-up of 13.4 months (range, 3.0-59.0 months), 184 patients in the XELOX arm and 176 in the EOX arm died. In the ITT set, mOS was 12.0 months (95% CI = 10.4-14.0 months) in the XELOX arm and 12.0 months (95% CI = 11.0-14.0 months) in the EOX arm (HR = 1.098; 95% CI = 0.893-1.350; P = 0.384; Figure 2C). In the PP set, mOS was 12.5 months (95% CI = 10.5-14.0 months) in the XELOX arm and 12.0 months (95% CI = 10.0-14.0 months) in the EOX arm (HR = 1.071; 95% CI = 0.867-1.323; P = 0.529; Figure 2D).

Subgroup analysis was performed to analyze the PFS and OS of the eligible patients according to ECOG status, locally advanced/metastasis, sex, age, primary tumor location, histologic grade, pathologic type, metastasis site, number of metastatic sites, and HER2 status (Figure 3, Supplementary Figures S1-3, Supplementary Table S2). Subgroup analysis showed that the EOX arm had longer OS than the XELOX arm in the HER2 unknown subgroup. This might be due to bias, as the HER2 unknown subgroup included both HER2-positive and -negative patients. However, there was no difference in OS between the two arms in the HER2-positive or -negative subgroup. There were no significant differences in PFS or OS in other subgroup analysis. Further exploratory analysis was performed in patients with poorly differentiated adenocarcinoma and liver metastases. The results showed that, in the ITT and PP sets, the EOX arm had a significantly longer mOS (P = 0.021 and P = 0.020, respectively) and a longer mPFS (P = 0.073, P = 0.057, respectively) than the XELOX arm (Supplementary Figure S4), while the clinicopathological characteristics between the two arms were balanced (Supplementary Table S3).

In the ITT set, the ORR was 39.0% in the XELOX arm and 47.4% in the EOX arm (P = 0.291) (Supplementary Table S4).

3.3 Safety

The most common AEs with a rate of ≥ 5% are summarized in Table 2. The rate of grade 3/4 AEs was 42.2% (90/213) in the XELOX arm and 72.5% (156/215) in the EOX arm (P = 0.001). Grade 3/4 white blood cell decrease, neutrophil decrease, platelet decrease, nausea, and vomiting occurred more often in the EOX arm than in the XELOX arm (all P < 0.05). One treatment-related death (lung infection) was observed in the EOX arm, and none was observed in the XELOX arm.

3.4 Quality of life

At baseline and at the beginning of cycles two, four, six, and eight, the rates of compliance to the QoL questionnaire were 86.9%, 75.3%, 68.1%, 48.8%, and 15.0% in the XELOX arm, and 83.2%, 81.8%, 74.1%, 45.9%, and 12.1% in the EOX arm, respectively. The QoL scores at baseline were similar between the two arms (P = 0.425). At the beginning of cycles two, four, six, and eight, QoL scores were significantly higher in the XELOX arm than in the EOX arm (Figure 4A), which were confirmed by using the LOCF method (Figure 4B) and Mean-value method (Figure 4C).

ETHICS APPROVAL AND CONSENT TO PARTICIPATE

This trial was approved by ethics committee at each participating site and the trial obtained patients’ consent to participate.

CONSENT FOR PUBLICATION

All authors have consent for publication.

CONFLICT OF INTERESTS

The authors declare no potential conflicts of interest.