Maintenance capecitabine after first-line platinum-based chemotherapy in advanced oesophagogastric adenocarcinoma: final analysis from the PLATFORM trial

PLATFORM is an adaptive phase II trial assessing maintenance therapies in advanced oesophagogastric adenocarcinoma (OGA). We evaluated maintenance capecitabine in patients with disease control after first-line chemotherapy.

HER2-negative patients with advanced OGA who had response or stable disease after 18 weeks of first-line chemotherapy were randomised (1:1) to surveillance or capecitabine. The primary endpoint was progression-free survival (PFS); secondary endpoints included overall survival (OS) and safety.

Between May 2015 and May 2024, 266 patients were randomised (129 surveillance, 137 capecitabine). Median follow up was 70.7 months. Capecitabine significantly improved PFS (HR 0.69; 95% CI 0.54–0.89; p = 0.002), with median PFS of 5.0 vs 2.8 months. One-year PFS rates were 19.9% vs 6.8%; and two-year rates 8.1% vs 4.3%. No OS difference was observed (median OS: 10.5 vs 10.0 months; HR 0.87; 95% CI 0.67–1.12; p = 0.143). One and two-year OS rates were similar (1-year: 44.1% vs 45.7%; 2-year: 18.8% vs 16.8%). Grade ≥3 adverse events were more frequent with capecitabine (46% vs 29%), with 21% experiencing grade 3 treatment related events.

Maintenance capecitabine significantly prolonged PFS compared to surveillance, meeting the primary endpoint and supporting its use to extend disease control in advanced OGA.

Gastric and oesophageal cancers are among the leading causes of cancer-related mortality world-wide, as most patients are diagnosed at an advanced or metastatic stage, where curative surgery is no longer an option [ 1 ]. Palliative treatment with fluoropyrimidine and platinum-based chemotherapy has improved survival for advanced oesophagogastric cancer, but overall outcomes remain poor, with median overall survival (OS) typically less than 12 months. The addition of targeted therapies such as trastuzumab for HER2-positive tumours, immune checkpoint inhibitors (ICI) targeting PD-L1, and more recently, zolbetuximab, to first-line chemotherapy has further extended median OS up to 18 months in selected populations [ 2 , 3 , 4 , 5 , 6 , 7 ]. However, even among patients who achieve an objective response to first-line treatment, disease progression is almost inevitable and approximately 30-50% of patients receive subsequent therapy [ 3 , 4 , 7 , 8 ]. The aggressive nature of advanced oesophagogastric adenocarcinoma (OGA) may limit the feasibility of a surveillance only approach after first-line therapy, as rapid clinical deterioration at progression may preclude timely delivery of further treatment.

In several cancers, including non-small cell lung cancer, ovarian, colorectal, breast, bladder and pancreatic cancer, maintenance therapy is an established approach that prolongs disease control and improves survival [ 9 , 10 , 11 , 12 , 13 , 14 , 15 , 16 ]. Approaches include de-escalated or ‘stop and go’ chemotherapy, which maintains disease suppression while minimising side effects, and switch maintenance, where a different agent with a new mechanism of action is introduced to delay resistance survival [ 17 ]. Targeted therapies also have an important role in maintenance treatment, and in advanced OGA, this has been demonstrated with trastuzumab and, more recently, ICIs and zolbetuximab.

The PLATFORM trial (PLAnning Treatment For Oesophago-gastric Cancer) was a multi-centre, adaptive, randomised phase II study evaluating maintenance therapies in patients with advanced, HER2-negative OGA who responded to or had stable disease following first-line chemotherapy. The original protocol included surveillance, capecitabine or durvalumab, with additional arms investigating rucaparib and capecitabine plus ramucirumab (cape-ram) added later. Each investigational arm was compared to active surveillance. Recruitment to the trial was completed in May 2024 once the accrual target for the capecitabine analysis was met. Recruitment to the durvalumab, rucaparib and cape-ram arms closed early due to loss of industry support [ 18 , 19 , 20 ]. The durvalumab and rucaparib arms previously met futility criteria, and neither demonstrated a significant progression-free survival (PFS) benefit compared with surveillance (durvalumab: hazard ratio (HR) 0.81; p = 0.13) or maintenance rucaparib (HR 0.68; p = 0.061) [ 18 , 19 ]. The maintenance cape-ram arm also closed early as the drug patent for ramucirumab was due to expire, leading to industry withdrawal. Cape-ram significantly prolonged PFS (HR 0.33; p p = 0.023) compared to surveillance, however this analysis was limited by small sample size [ 20 ].

Here, we present primary results comparing maintenance capecitabine with surveillance in HER2- negative advanced OGA patients following disease control with first-line chemotherapy.

Eligible patients were adults (aged ≥18 years) with histologically confirmed, inoperable locally advanced or metastatic adenocarcinoma of the oesophagus, gastro-oesophageal junction or stomach, and who had achieved SD or better on CT scan after first-line fluoropyrimidine-platinum chemotherapy. The initial protocol mandated completion of at least six cycles of investigator’s choice first-line fluoropyrimidine chemotherapy. This was then limited to 18 weeks and the specific regimens mandated were capecitabine and oxaliplatin (CAPOX), cisplatin and capecitabine (CX) or 5-fluorouracil and oxaliplatin (FOLFOX). Patients were required to have HER2- negative tumours, performance status (PS) 0-2 and adequate organ function. Measurable disease was not required for trial entry. PD-L1 and mismatch repair (MMR) status were not an eligibility criterion as it was not standard of care when the protocol was established. However, this was assessed in the experiment arm with durvaluamb [ 19 ].

PLATFORM is an open-label, multicentre, randomised phase II trial that recruited patients from 46 centres across the United Kingdom. The adaptive design allowed for early closure of ineffective treatment arms and the addition of new investigational arms. Patients were registered before or during first-line chemotherapy. Those who achieved disease control after 18 weeks of treatment and met eligibility criteria were randomised to surveillance (control arm) or one of the maintenance arms. The initial randomisation included surveillance, capecitabine or durvalumab in the original protocol design. Rucaparib and capecitabine plus ramucirumab arms were later added, resulting in a final randomisation of 1:1:1:1:1 across the five arms. (Supplementary fig. 1 ). Randomisation was performed at the Institute for Cancer Research Clinical Trials and Statistics Unit by random permuted blocks.

This analysis included patients randomised to surveillance contemporaneously with capecitabine. Surveillance visits were undertaken every 28 days, and patients assigned to capecitabine received continuous dose capecitabine 1250 mg/m 2 /day of a 21-day cycle to progression, death, or toxicity.

The primary endpoint was progression-free survival (PFS), defined as time from randomisation to radiological (per RECIST 1.1) or clinical progression or death from any cause. Secondary endpoints included OS, time from randomisation to death from any cause, objective response rate (ORR), progression-free rate (PFR) and safety. Randomisation was stratified by region, disease extent (locally advanced versus metastatic disease) and ECOG performance status (0 versus 1/2). CT imaging was performed at baseline and every 12 weeks. Adverse events (AEs) were recorded from randomisation to 30 days post-treatment and graded per National Cancer Institute Common Terminology Criteria for Adverse Events version 4.0.

This was an investigator-initiated, academically sponsored trial (Royal Marsden NHS Foundation Trust). Ethics approval was obtained from the UK National Research Ethics Committee London Southeast (14/LO/1206) and the Medicines and Healthcare Products Regulatory Agency (Eudra CT Number: 2014-002169-30). Written informed consent was obtained from all patients. Trial oversight was provided by the sponsor, investigators, research ethics committee and an independent data monitoring committee (IDMC). The trial is registered at ClinicalTrials.gov (NCT02678182).

To detect a PFS hazard ratio (HR) of 0.7 with 80% power and a one-sided alpha of 0.025, 154 patients per arm were required, assuming a 10% dropout rate and median PFS of 6 months in the surveillance arm (data from REAL-2) [ 21 ]. A total of 248 PFS events were needed for definitive analysis, assuming a 3-year accrual and maximum 4-year follow-up. A futility analysis based on 12-week PFR was triggered once 61 evaluable patients per interventional arm underwent a restaging CT scan post-randomisation. Data were analysed using Stata version 18.5 (StataCorp, College Station, TX).

Between May 2015 and May 2024, 1413 patients were registered across 46 participating centres. Of these, 1118 (79%) completed first-line chemotherapy, and 494 (44%) patients were randomised across all treatment arms. The most common reason for non-randomisation was disease progression (44.5%). During this period, 129 patients were allocated to surveillance and 137 patients to the capecitabine arm, forming the intention-to-treat (ITT) population ( n = 266). One patient from each group withdrew after randomisation but before starting treatment, resulting in a safety population of 128 surveillance and 136 capecitabine patients.

After starting treatment/surveillance, patient withdrawal occurred in 15 patients in the surveillance arm (12 full withdrawals by patient choice, and 3 partial withdrawals) and 18 patients in the capecitabine arm (4 full withdrawals by patient choice, and 14 partial withdrawals). All withdrawn patients were retained in the primary endpoint analysis of PFS. However, for patients who withdrew full consent, no additional on study data, including OS information, could be collected from date of withdrawal. One surveillance patient was lost to follow up while none were lost in the capecitabine arm. Figure 1 shows the CONSORT diagram.

Table 1 shows baseline characteristics which are similar between both groups. Most patients achieved stable disease following first-line chemotherapy, and the proportion of patients achieving a complete or partial response was comparable between the two arms. At data cut-off (17 April 2025), two surveillance and five patients in the capecitabine arm remained on treatment. The most common reason for discontinuation was disease progression (88% versus 84%, respectively) and 7% of patients assigned to capecitabine discontinued due to toxicity. The median duration of capecitabine therapy was 3.6 months (range: 0.1–57 months), with a median of five treatment cycles (range: 1–76). Dose reductions were required in 61 patients, primarily due to toxicity (77%). Supplementary table 1 outlines post-progression therapies. More patients in the surveillance group received overall further subsequent therapy (73% vs. 57%), and platinum rechallenge was more frequent in the surveillance group (29% vs. 16%).

After a median follow-up of 70.7 months (95% confidence interval (CI) 39.7—not estimable), 248 PFS events had occurred, meeting the required number for the primary analysis. Of these events, 242 were due to disease progression and six due to death without prior progression. Progression occurred in 91% of patients in both the surveillance (117/129) and capecitabine (125/137) arms. Death occurred in 86% (111/129) and 91% (124/137) of patients, respectively.