A phase II randomized trial of gemcitabine plus cisplatin (GP) versus gemcitabine plus carboplatin (GC) as the first-line treatment of patients with metastatic triple-negative breast cancer.

Introduction
Triple-negative breast cancer (TNBC) is a highly aggressive and therapeutically challenging subtype compared with other subtypes.1 Although immunotherapy,2,3 and poly (ADP-ribose) polymerase (PARP) inhibitors4 have demonstrated clinical activity, these options are limited to biomarker-defined subgroups. The majority of patients in the first-line setting remain ineligible for immunotherapy, however, either due to programmed death-ligand 1 (PD-L1) combined positive score (CPS) <10 or prior exposure to immune checkpoint inhibitors (ICIs) in early-stage disease. With the widespread adoption of neoadjuvant immunotherapy, the population of ICI-pretreated patients is steadily increasing, further necessitating effective non-immunotherapy treatment backbones. Recent trials such as ASCENT-035 and TROPION-Breast026 have demonstrated the superior efficacy of single-agent antibody–drug conjugates (ADCs) compared with standard chemotherapy in ICI-ineligible patients. Nevertheless, these agents have not yet been approved for first-line use. Furthermore, even following potential future approvals, access would remain limited in many healthcare systems due to regulatory and reimbursement barriers and economic constraints. As such, chemotherapy remains the essential first-line treatment of the majority of patients in many clinical scenarios.
Increasing evidence highlights the pivotal role of platinum-based chemotherapy in treating TNBC. A multiple Bayesian network meta-analysis involving 125 first-line clinical trials and 37 812 patients with metastatic TNBC (mTNBC) confirmed that platinum-based regimens ranked top for all endpoints, including objective response rate (ORR), progression-free survival (PFS), and overall survival (OS)—as determined by the surface under the cumulative ranking curve values.7
Gemcitabine plus carboplatin (GC) is one of the classic chemotherapeutic regimens for patients with mTNBC, with a median PFS of 4.6 months as first-line treatment, recommended by National Comprehensive Cancer Network (NCCN) guidelines as an option for mTNBC.8 In recent years, immunotherapy has transformed the treatment landscape for TNBC. The phase III KEYNOTE-355 trial has established the combination of pembrolizumab and chemotherapy (nab-paclitaxel, paclitaxel, or GC) as the standard of care for first-line treatment in PD-L1-positive mTNBC. Subgroup analysis of the 465 patients treated with GC, however, showed that adding pembrolizumab did not confer a significant survival benefit [7.4 versus 7.4 months, hazard ratio (HR) 0.93, 95% confidence interval (CI) 0.74-1.16]. Although no definitive conclusions could be drawn from subgroup results, they reflect the robust efficacy of the GC doublet, which potentially leaves limited room for further incremental benefit from the addition of immunotherapy. Since only ∼40% of TNBC patients meet the PD-L1 CPS ≥10 threshold, platinum-based chemotherapy remains an essential component of treatment. These findings indirectly support the clinical value of the GC regimen in the first-line mTNBC setting.9
The TBCRC009 study showed that the ORR of cisplatin monotherapy in the treatment of mTNBC was higher than that of carboplatin, but there was no statistical difference (32.6% versus 18.7%).10 Our previous phase III CBCSG006 trial demonstrated that first-line cisplatin plus gemcitabine (GP) significantly prolonged PFS compared with paclitaxel plus gemcitabine among patients with mTNBC (7.73 versus 6.47 months, P = 0.009).11 Thus, the GP regimen has been recommended in the guidelines of the Chinese Breast Cancer Society and the Chinese Society of Clinical Oncology.12,13
Platinum-based chemotherapy remains a cornerstone of first-line treatment of mTNBC. Whether cisplatin confers superior benefit over carboplatin in combination with gemcitabine, however, is unclear.
To address this, a prospective, randomized phase II trial (the GPGC study) was conducted to compare the efficacy and safety of GP versus GC as the first-line treatment of patients with mTNBC, aiming to provide evidence-based guidance for clinical decision-making.

Patients and methods

Study design and participants
This open-label, single-center, phase II, randomized clinical trial was conducted at Fudan University Shanghai Cancer Center (FUSCC). Details of the protocol are described in Supplementary Material S1, available at https://doi.org/10.1016/j.esmoop.2026.106889. Eligible participants had histologically confirmed mTNBC, at least one measurable lesion according to Response Evaluation Criteria in Solid Tumors (RECIST) 1.1, Eastern Cooperative Oncology Group (ECOG) performance status score of 0 or 1, adequate organ function, a life expectancy of at least 12 weeks, and a 6-month gap since their last neoadjuvant or adjuvant chemotherapy dose. No prior chemotherapy for metastatic breast cancer is permitted. Key exclusion criterion was symptomatic central nervous system metastases.
The study was conducted in accordance with the Declaration of Helsinki. The study protocol, statistical analysis plan, and informed consent form were approved by the ethics committee of FUSCC (IRB: 1412142-14-1609B). All patients provided written informed consent. This study was registered with ClinicalTrials.gov Identifier: NCT 02341911. This article was prepared in accordance with the Consolidated Standards of Reporting Trials (CONSORT) reporting guideline (Supplementary Material S2, available at https://doi.org/10.1016/j.esmoop.2026.106889).

Randomization and procedures
Patients were randomized 1 : 1 using block randomization, stratified by visceral metastasis (yes versus no) and the number of metastases (1 versus 2 versus ≥3).
Patients in the GP group received gemcitabine [1250 mg/m2 intravenously (i.v.) over 30 min on days 1 and 8] plus cisplatin (75 mg/m2 i.v. over 120 min on day 1). To prevent renal toxicity, cisplatin was administered with a mandatory 3-day hydration protocol (minimum 2500 ml/day). For antiemetic prophylaxis, a standard regimen consisted of a 5-HT antagonist, aprepitant, and dexamethasone on days 1-3. Patients in the GC group received gemcitabine (1000 mg/m2 i.v. on days 1 and 8) plus carboplatin [area under the curve (AUC) 2 i.v. on days 1 and 8]. Both regimens were repeated every 21 days until disease progression, intolerable toxicity, or treatment delay for >14 days.
Dose modifications were regulated by protocol-specified toxicity criteria (Supplementary Material S1, Protocol Section 7, available at https://doi.org/10.1016/j.esmoop.2026.106889). Tumor assessment of evaluable lesions was carried out by computed tomography scanning or magnetic resonance imaging at baseline, every two cycles during treatment. Tumor response was evaluated using RECIST v1.1. The adverse events (AEs) were evaluated from the date of informed consent to 28 days after the last dose of chemotherapy.

Outcomes
The primary endpoint was investigator-assessed PFS as per RECIST (version 1.1), which was defined as the time from randomization to objective tumor progression or death from any cause, whichever occurred first. Secondary endpoints included OS, ORR, and safety. Safety was assessed according to National Cancer Institute Common Terminology Criteria for Adverse Events (NCI CTCAE version 4.0).

Statistical analysis
The sample size was determined based on the results of two phase III clinical trials: the median PFS for patients receiving GC and GP as first-line treatment of mTNBC was 4.6 and 7.7 months, respectively.11,14 We hypothesized that GP would be superior to GC in terms of efficacy. To detect an improvement of median PFS from 4.6 months to 7.7 months, with 80% power and a type I error of 0.05 (24 months enrollment duration, 12 months follow-up duration after enrollment), 136 patients would be required. Considering a drop-out rate of 10%, a total of 150 patients were planned to be enrolled.
Efficacy was analyzed in the intention-to-treat (ITT) population, which included all randomly assigned participants. Safety was analyzed in all patients with safety records who received at least one dose of the study drug.
PFS and OS were estimated using the Kaplan–Meier method and compared using the log-rank test. HRs and corresponding 95% CIs were estimated using the Cox proportional-hazards regression model. Stratified HR and 95% CI were calculated using the factors applied for randomization. Post hoc subgroup analyses were carried out using a stratified Cox proportional-hazards model with two-sided 95% CIs. The ORR was compared using the chi-square test. Dose intensity is the amount of drug (mg/m2) delivered to a patient in a week of treatment.15 Safety was assessed in all patients who received at least one dose of the trial treatment. The grade of AEs during the trial treatment was reported. All statistical tests were two-sided. Differences were considered statistically significant if P < 0.05. Statistical analyses were done with SPSS (version 25.0.0.1) and R (version 4.2.2).

Results

Patients
Between 5 January 2015 and 6 December 2022, a total of 150 patients were enrolled and randomized to receive GP (n = 75) or GC (n = 75) (Figure 1). All 150 patients were included in the ITT population. In the GP group, 1 patient did not initiate treatment after randomization; therefore, 149 patients were included in the safety analysis set. Patient demographics and baseline disease characteristics were well balanced between the treatment groups (Table 1). 69.3% of patients in the GP group and 60.0% in the GC group had visceral metastasis. Most patients had received anthracycline- or taxane-based adjuvant or neoadjuvant chemotherapy. Nine (12.0%) patients in the GP group and seven (9.3%) in the GC group received capecitabine as adjuvant treatment. Six patients (4.0%) had received platinum-containing therapy in the early-stage setting: five (6.7%) in the GP group and one patient (1.3%) in the GC group (P = 0.21). Baseline renal function was also comparable between the two arms; the mean baseline creatinine level was 51.64 μmol/l in the GP arm and 52.74 μmol/l in the GC arm (P = 0.53).

Treatment exposure
In both groups, patients received a median of six cycles of treatment (range: 0-16 for the GP group and 1-20 for the GC group). A total of 30 (40.0%) patients in the GP group and 38 (50.7%) in the GC group experienced dose reduction. In the GP group, the median dose intensity of cisplatin was 21.6 mg/m2/week, with a relative dose intensity (RDI) of 86.4%. The median dose intensity for gemcitabine was 639.1 mg/m2/week, with an RDI of 76.7%. In the GC group, the median dose intensity of carboplatin was 85.7 mg/m2/week, with an RDI of 80.6%, while the median dose intensity of gemcitabine was 533.2 mg/m2/week, with an RDI of 80.0% (Supplementary Material S3 and Table S1, available at https://doi.org/10.1016/j.esmoop.2026.106889).

Efficacy
The median follow-up time was 57.1 months (interquartile range 42.0-85.7 months). At the data cut-off (16 April 2024), PFS events had been recorded in 55 patients (73.3%) in the GP group and 59 (78.7%) in the GC group. PFS was numerically longer in the GP group than in the GC group, though not significantly [median 7.8 months (95% CI 7.2-8.4 months) versus 7.0 months (95% CI 6.6-7.5 months); HR 0.94, 95% CI 0.65-1.36, P = 0.74; stratified HR 0.86, 95% CI 0.59-1.25, P = 0.43] (Figure 2A).
For subsequent medical treatments after disease progression (Supplementary Material S3 and Table S2, available at https://doi.org/10.1016/j.esmoop.2026.106889), 34 (61.8%) of 55 patients in the GP group and 41 (69.5%) of 59 patients in the GC group received subsequent systemic treatments. A higher proportion of patients in the GC group received combination regimens as second-line treatment compared with the GP group (30.5% versus 16.4%, P = 0.076). The most commonly used combination regimen was vinorelbine-based chemotherapy. Additionally, anti-angiogenic therapies were administered in six patients (10.2%) in the GC group and one patient (1.8%) in the GP group.
Exploratory post hoc subgroup analyses of PFS are shown in Figure 3. The results were consistent across the main clinical subgroups.
In the ITT population, the ORR was 49.3% in the GP group and 41.3% in the GC group (OR 1.38, 95% CI 0.73-2.63, P = 0.33). Among patients assessable for response, the ORR was 58.7% and 48.4%, respectively (OR 1.52, 95% CI 0.75-3.05, P = 0.25, Table 2).
At the time of the data cut-off, 56 patients (74.7%) in the GP group and 59 (78.7%) in the GC group had died. OS was similar between the two groups, with a stratified HR of 1.05 (95% CI 0.73-1.52; P = 0.79; HR 1.13, 95% CI 0.78-1.64, P = 0.51). The median OS was 20.3 months (95% CI 17.6-23.1 months) for the GP group and 19.3 months (95% CI 13.9-24.6 months) for the GC group (Figure 2B).

Safety
All 149 patients in the safety population experienced at least one AE (Table 3). Hematological AEs were the most common, with similar incidences in both groups. Leukopenia occurred in 70 of 74 patients in the GP group (94.6%) and 67 of 75 patients in the GC group (89.3%) (P = 0.24); neutropenia in 70 (94.6%) versus 64 (85.3%) (P = 0.06); anemia in 68 (91.9%) versus 64 (85.3%) (P = 0.21), and thrombocytopenia in 62 (83.8%) versus 57 (76.0%) (P = 0.24). Grade 3-4 hematological AEs were slightly more common in the GC group, including neutropenia (48.0% versus 44.6%), thrombocytopenia (45.3% versus 44.6%), leukopenia (42.7% versus 39.2%), and anemia (26.7% versus 20.3%). Non-hematological AEs were reported more often with GP, including nausea (67.6% versus 48.0%, P = 0.016), vomiting (40.5% versus 26.7%, P = 0.07), anorexia (36.5% versus 29.3%, P = 0.35), peripheral neuropathy (17.6% versus 6.7%, P = 0.041), tinnitus (10.8% versus 4.0%, P = 0.13), and hearing toxicity (6.8% versus 1.3%, P = 0.12). Rash, however, was more common in the GC group (12.0% versus 5.4%, P = 0.25).
Laboratory abnormalities also showed group-specific patterns. Elevated alanine aminotransferase (ALT) (49.3% versus 37.8%, P = 0.16) and aspartate aminotransferase (AST) (41.3% versus 31.1%, P = 0.19) were more frequent in the GC group. By contrast, renal and electrolyte disturbances were more frequent with GP, including increased creatinine (13.5% versus 1.3%, P = 0.005), hypomagnesemia (44.6% versus 25.3%, P = 0.01), hyponatremia (17.6% versus 5.3%, P = 0.02), and hypophosphatemia (21.6% versus 8.0%, P = 0.02). No treatment-related deaths were reported.
Exploratory retrospective analyses were conducted for biomarkers and prior platinum exposure. Due to antigen degradation in archival slides, PD-L1 testing was not feasible. Germline BRCA testing was available for a limited subset (n = 48), and the descriptive results are provided in Supplementary Material S3 and Table S3, available at https://doi.org/10.1016/j.esmoop.2026.106889. Additionally, six patients (4.0%) had received prior platinum-containing therapy in the (neo)adjuvant setting; sensitivity analyses excluding these individuals yielded results consistent with the primary ITT analysis in Supplementary Material S3 and Tables S4 and S5, available at https://doi.org/10.1016/j.esmoop.2026.106889.

Discussion
Despite recent therapeutic advances, platinum-based chemotherapy remains an important first-line regimen for mTNBC, particularly for patients ineligible for immunotherapy. Importantly, even in the ASCENT-03 trial, 44% of investigators selected GC as their preferred chemotherapy backbone before randomization.5 This physician’s preference highlights the continued reliance on platinum-based doublets and supports the continued refinement of such regimens. This randomized controlled trial is, to our knowledge, the first to directly compare GP with GC as the first-line treatment of mTNBC. GP was associated with numerically longer PFS and higher ORR compared with GC, although these differences were not statistically significant. No difference in OS was observed between groups. The toxicity profiles differed between the two regimens but remained within manageable limits.
The efficacy of GP observed in this study is consistent with previous trials conducted at our center, including CBCSG006 and GAP, in which median PFS consistently ranged from 7.2 to 7.8 months, with reproducible efficacy across multiple cohorts11,16,17 (Supplementary Material S3 and Table S6, available at https://doi.org/10.1016/j.esmoop.2026.106889). Reported toxicities, including myelosuppression, gastrointestinal effects, nephrotoxicity, and ototoxicity, were also in line with prior findings.
For GC, our observed median PFS of 7.0 months and OS of 19.3 months compared favorably with historical studies in the same setting. In the tnAcity trial, GC achieved a median PFS of 6.0 and OS of 12.6 months.18 Compared with tnAcity, our study enrolled younger patients (median age 53 versus 59 years) and had a higher proportion with ECOG PS 1 (86.7% versus 33%), but fewer patients with liver metastases (21.3% versus 35%, Supplementary Material S3 and Table S7, available at https://doi.org/10.1016/j.esmoop.2026.106889). Rates of dose reduction due to toxicity (50.7% versus 52%) and grade ≥3 anemia (26.7% versus 27%) were similar across studies, while grade ≥3 thrombocytopenia was more frequent in our cohort (45.3% versus 28%). These differences may be partly attributable to ethnic variation, as all patients in our study were Chinese, whereas 82% of tnAcity participants were white. The phase III trial by O’Shaughnessy et al.14 also reported a median PFS of 4.6 months and OS of 12.6 months for the GC arm, though direct comparison is limited due to the absence of baseline characteristics for patients receiving first-line GC. In the KEYNOTE-355 study, GC achieved a median PFS of 7.4 months in the ITT population.9 In the ASCENT-03 trial, which evaluated sacituzumab govitecan versus chemotherapy in the first-line setting for ICI-ineligible mTNBC, the median PFS reached 8.1 months in the GC subgroup.5 This represents the longest reported PFS for this regimen to date and underscores the continued foundational relevance of the GC doublet in contemporary clinical practice. Subgroup characteristics of these trials were not disclosed, however, making it unclear whether differences in disease burden accounted for this result.
The question of whether cisplatin and carboplatin are therapeutically equivalent has long been debated. Several randomized trials across tumor types have yielded inconsistent conclusions.19 In advanced non-small-cell lung cancer (NSCLC), a phase III trial comparing paclitaxel plus cisplatin (TP) versus paclitaxel plus carboplatin (TC) found similar ORR but significantly longer OS with TP (median OS 9.8 versus 8.5 months, HR 1.2, 95% CI 1.03-1.40), with no significant difference in quality-of-life scores.20 Another randomized phase II trial comparing GP and GC in metastatic NSCLC found higher ORR with GP, but no OS difference, suggesting carboplatin may be a feasible alternative.21 A meta-analysis of nine first-line NSCLC trials involving 2968 patients showed a higher ORR with cisplatin (30% versus 24%, P < 0.01), and subgroup analysis suggested survival benefits in non-squamous histology and when combined with third-generation agents such as gemcitabine and taxanes.22 In small-cell lung cancer, large-scale meta-analyses have shown no significant efficacy difference between cisplatin- and carboplatin-based regimens.23 In advanced ovarian cancer, randomized trials have found similar efficacy between TP and TC regimens, though carboplatin was associated with better tolerability and quality of life.24
In TNBC, a retrospective study of 145 early-stage patients receiving platinum-based neoadjuvant chemotherapy showed no significant difference in pathologic complete response (pCR), disease-free survival (DFS), or OS between TP and TC regimens.25 In metastatic breast cancer, a randomized phase II trial comparing biweekly GT, GP, and GC regimens showed no significant differences in survival. Only a subset of patients had TNBC, however, and the regimens did not follow the standard 3-week schedule. In that study, the median PFS for GP and GC was 4.8 months (95% CI 3.7-8.1 months) and 4.3 months (95% CI 3.7 months-not reached), respectively.26 A retrospective cohort study at FUSCC reported significantly longer median PFS in mTNBC patients receiving cisplatin-based first-line therapy compared with other platinum agents (8.0 versus 4.3 months, P = 0.03).27 Our current trial confirms that GP and GC yield comparable efficacy in the first-line treatment of mTNBC.
In terms of safety, the toxicity profiles were consistent with those previously reported. Non-hematological AEs were more common in the GP group, including nausea (67.6% versus 48.0%, P = 0.016), vomiting (40.5% versus 26.7%, P = 0.07), anorexia (36.5% versus 29.3%, P = 0.35), peripheral neuropathy (17.6% versus 6.7%, P = 0.041), tinnitus (10.8% versus 4.0%, P = 0.13), and hearing toxicity (6.8% versus 1.3%, P = 0.12). Most cases were grades 1-2. Laboratory abnormalities reflected regimen-specific patterns: abnormalities in liver function were more common in the GC group while renal dysfunction and electrolyte disturbances (e.g. increased creatinine, hypomagnesemia, hyponatremia) were more common in GP. No treatment-related deaths occurred.
The present study has some limitations that warrant consideration. First, in recent years, immunotherapy has significantly reshaped the treatment landscape for TNBC, yet our findings cannot be directly extrapolated to platinum-based chemoimmunotherapy settings. Both GP and GC regimens have shown feasibility and promising efficacy when combined with programmed cell death protein 1 (PD-1) inhibitors. In the ITT population of the KEYNOTE-355 trial, the GC regimen with pembrolizumab achieved a median PFS of 7.4 months as first-line treatment.9 Similarly, a phase I trial assessing GP combined with pucotenlimab, a PD-1 inhibitor, as first-line therapy for mTNBC demonstrated a median PFS of 9.0 months.28 Mechanistically, platinum agents are known to induce immunogenic cell death and may synergize with immune checkpoint blockade. Cisplatin, in particular, has been shown to up-regulate major histocompatibility complex (MHC) class I expression on tumor and antigen-presenting cells, recruit and expand effector immune cells, enhance cytolytic activity, and reduce the expression of immunosuppressive mediators in the tumor microenvironment.29,30 In contrast, the immunomodulatory properties of carboplatin are increasingly recognized,31 representing a growing area of interest. Second, although our sensitivity analysis indicated that the small number of patients with prior (neo)adjuvant platinum exposure (4.0%) did not affect the primary efficacy outcomes, this low prevalence reflects the treatment standards at the time of study initiation. In the current era, platinum-based regimens are increasingly used in the neoadjuvant setting for TNBC. Due to the limited number of such patients in our cohort, we were unable to draw definitive conclusions about how prior platinum exposure in the early stage might affect the efficacy of first-line platinum-based doublets in the metastatic setting. This remains a critical area for future investigation as treatment paradigms continue to evolve. Furthermore, biomarker characterization was limited. PD-L1 testing was not prospectively incorporated, and retrospective assessment was not feasible due to poor antigen preservation in archival specimens. Additionally, germline BRCA testing was available only in a small subset of patients and was carried out using non-standardized methodologies, precluding robust biomarker-based subgroup analyses. These factors limit interpretation in the context of biomarker-driven treatment paradigms. Lastly, this study did not include quality-of-life assessments, so it cannot compare patient-reported outcomes or tolerability differences between cisplatin and carboplatin.

Conclusion
GP and GC demonstrated comparable efficacy as first-line regimens for mTNBC, although numerically longer PFS and a higher ORR were observed in GP group. The safety profiles of the two combinations were different, providing evidence to support individualized regimen selection in clinical practice.