Effect of adjuvant carboplatin intensified chemotherapy versus standard chemotherapy on survival in women with high risk, early stage, triple negative breast cancer (CITRINE): randomised, open label phase 3 trial Authors: Yin Liu, Yue Gong, Xiu-Zhi Zhu, Guang-Yu Liu, Ke-Da Yu, Fan Yang, Li Chen, Min He, Zhen Hu, Can-Ming Chen, A-Yong Cao, Jun-Jie Li, Yi-Feng Hou, Gen-Hong Di, Jiong Wu, Yi-Zhou Jiang, Lei Fan, Zhong-Hua Wang, Zhi-Ming Shao Published: 2025 ABSTRACT AbstractObjectiveTo evaluate the efficacy and safety of epirubicin and cyclophosphamide followed by weekly paclitaxel with or without carboplatin as adjuvant therapy for patients with high risk, early stage, triple negative breast cancer.DesignRandomised, open label, phase 3 trial.SettingFudan University Shanghai Cancer Center in China between March 2020 and March 2022.ParticipantsFemale patients with operable high risk triple negative breast cancer (defined as either regional node positive or node negative with a Ki-67 labelling index of ≥50%) after definitive surgery.InterventionsPatients were randomised in a 1:1 ratio into either the carboplatin arm (four cycles of two weekly epirubicin and cyclophosphamide followed by four cycles of weekly paclitaxel combined with carboplatin) or the control arm (four cycles of three weekly or two weekly epirubicin and cyclophosphamide followed by four cycles of weekly paclitaxel).Main outcome measuresThe primary endpoint was disease-free survival in the intention-to-treat population. Secondary endpoints included recurrence-free survival, distant disease-free survival, overall survival, and safety.ResultsOf the 808 enrolled patients, 807 received study treatment. At a median follow-up of 44.7 months, estimated percentages of patients who would be disease-free at three years were 92.3% in the carboplatin arm and 85.8% in the control arm (hazard ratio 0.64, 95% confidence interval (CI) 0.43 to 0.95; P=0.03). However, the assessment of the proportional hazards assumption showed that it was violated (P=0.02). A piecewise hazard model showed that the hazard ratio changed over time: the hazard ratio was 0.31 (95% CI 0.13 to 0.73) for 0-12 months, 0.65 (0.39 to 1.09) for 12-36 months, and 1.98 (0.69 to 5.69) for >36 months. For secondary endpoints, the carboplatin arm was associated with improved outcomes in three year recurrence-free survival (93.8% v 88.3%; hazard ratio 0.59, 95% CI 0.37 to 0.93; P=0.02), three year distant disease-free survival (94.8% v 89.8%; hazard ratio 0.61, 0.37 to 0.98; P=0.04), and three year overall survival (98.0% v 94.0%; hazard ratio 0.41, 0.20 to 0.83; P=0.01). The incidence of grade 3-4 treatment related adverse events was 66.7% (269 of 403 patients) in the carboplatin arm and 55.0% (222 of 404 patients) in the control arm. No treatment related deaths occurred.ConclusionsThe addition of carboplatin to adjuvant anthracycline/taxane based chemotherapy significantly improved survival outcomes in patients with high risk, early stage, triple negative breast cancer, driven by reduction of early recurrence risk without new safety concerns.Trial registrationClinicalTrials.gov NCT04296175. Introduction Triple negative breast cancer, characterised by the absence of oestrogen receptor, progesterone receptor, and human epidermal growth factor receptor 2 (HER2), accounts for 15-20% of all invasive breast cancers and tends to have a higher risk of early recurrence than other subtypes.1 2 Although neoadjuvant treatment is typically recommended for patients with clinical stage II or III triple negative breast cancer, one observational study found that more than half of such patients still received upfront surgery.3 Moreover, a subset of patients who cannot be conclusively classified as stage II–III until after surgical intervention inevitably exists. Therefore, this patient subgroup warrants the identification of an optimal treatment regimen. The development of adjuvant therapies for early stage triple negative breast cancer has been a protracted and resource intensive process.4 The Eastern Cooperative Oncology Group (ECOG) E1199 study optimised the administration of taxanes and suggested that weekly paclitaxel, after anthracyclines, improves disease-free survival and overall survival in patients with triple negative breast cancer.5 6 The CALGB 9741 study and GIM-2 study confirmed that dose dense regimens further improved survival compared with the traditional three weekly regimen.7 8 Additionally, a meta-analysis by the Early Breast Cancer Trialists’ Collaborative Group showed that the dose dense two weekly regimen may reduce recurrence rates and mortality risk.9 Consequently, the National Comprehensive Cancer Network (NCCN) guidelines recommend preferred regimens for patients with HER2 negative breast cancer, including triple negative breast cancer, of dose dense anthracyclines followed by dose dense or weekly paclitaxel. Although the anthracycline-taxane sequential chemotherapy is the established standard for adjuvant treatment in early stage triple negative breast cancer, with a validated risk-benefit profile, approximately 30-40% of patients at high risk still experience disease recurrence and metastasis.5 6 10 11 12 Therefore, optimising adjuvant intensification strategies for patients with high risk triple negative breast cancer is a pressing concern. Platinum agents are cytotoxic, DNA damaging compounds leading to DNA strand breaks and cell apoptosis; this mechanism of action makes them especially active in cancer cells with DNA repair deficiency, such as those with homologous recombination deficiency.13 14 Many clinical trials have shown that platinum based agents can increase the pathological complete response rate to neoadjuvant therapy in patients with triple negative breast cancer. Nevertheless, significant controversy remains regarding whether platinum agents can improve patients’ long term survival and the association between the efficacy of platinum agents and the germline BRCA mutation status.15 16 17 18 19 20 21 22 However, the role of these agents in the adjuvant setting is still underexplored and supported by limited evidence.23 Questions remain about whether the incorporation of platinum drugs into adjuvant therapy for triple negative breast cancer confers survival benefits, particularly when combined with an optimal regimen comprising anthracyclines followed by weekly paclitaxel. In this context, we did a phase 3 trial (carboplatin intensified chemotherapy for triple negative breast cancer or CITRINE) to evaluate the efficacy and safety of epirubicin and cyclophosphamide, followed by weekly paclitaxel, with or without carboplatin, as adjuvant therapy for patients with high risk, early stage, triple negative breast cancer. Here, we present the primary findings of this trial after a median follow-up time of 44.7 months. Methods Study design and patients This was a randomised, open label, phase 3, two arm controlled trial (NCT04296175) designed to compare the efficacy and safety of a carboplatin intensified, dose dense regimen versus standard regimens in adjuvant therapy for high risk triple negative breast cancer. The study enrolled female patients aged 18-70 years with newly diagnosed, operable, unilateral invasive triple negative breast cancer who had undergone primary surgery with clear margins. Eligible patients were either pathologically confirmed as regional node positive or node negative with a Ki-67 index of at least 50% as an additional high risk feature. The status of oestrogen receptor, progesterone receptor, and HER2, as well as Ki-67 index, was centrally determined through immunohistochemical analysis at the Fudan University Shanghai Cancer Center. Oestrogen receptor/progesterone receptor negative status was defined as nuclear staining in <1% of tumour cells, and HER2 negative status was defined as an immunohistochemical score of 0/1 or a score of 2 without HER2 amplification, following the guidelines of the American Society of Clinical Oncology/College of American Pathologists.24 25 Inclusion criteria required an ECOG performance status of 0 or 1, normal organ function, and readiness for randomisation. We excluded patients if they had received preoperative anticancer therapy, had distant metastases, or were pregnant or breastfeeding. Detailed inclusion and exclusion criteria are outlined in the protocol (supplementary materials). The trial was conducted in accordance with Good Clinical Practice guidelines and the Declaration of Helsinki and was approved by the independent ethics committee at the participating institution. All participants gave written informed consent before their enrolment in the study. The trial was reported according to the CONSORT 2025 statement.26 Randomisation and masking Patients were assigned and randomised (1:1) to the carboplatin arm (two weekly epirubicin and cyclophosphamide followed by weekly paclitaxel combined with carboplatin) or the control arm (three weekly epirubicin and cyclophosphamide followed by weekly paclitaxel or two weekly epirubicin and cyclophosphamide followed by weekly paclitaxel, as determined by the investigator). Randomisation was done using a block randomisation method with a block size of eight, generated through a computer based allocation sequence created by an independent statistician. Investigators submitted randomisation forms to the Clinical Research Coordination Office at Fudan University Shanghai Cancer Center (Shanghai, China). Following verification of the patient’s eligibility on the basis of the inclusion and exclusion criteria, the study coordinator provided treatment arm assignments. Neither patients nor investigators were blinded to the treatment assignments, whereas all outcome assessors, data collectors, and analysts were blinded to the allocation. Procedures Patients in the carboplatin arm received four cycles of epirubicin 90 mg/m2 and cyclophosphamide 600 mg/m2 on day 1 every two weeks, sequentially followed by four cycles of paclitaxel 80 mg/m2 and carboplatin (area under the curve=2) on days 1, 8, and 15 of each 28 day cycle. Patients in the control arm received four cycles of epirubicin 90 mg/m2 and cyclophosphamide 600 mg/m2 on day 1 every two weeks or three weeks, sequentially followed by four cycles of paclitaxel 80 mg/m2 on days 1, 8, and 15 of each 21 day cycle. Treatment was discontinued in cases of disease progression, patient withdrawal, or unacceptable toxicity or at the investigator’s discretion. Guidelines for dose modification, premedication, and supportive care were outlined in the study protocol. Every patient assigned to dose dense anthracyclines received primary prophylaxis with pegylated recombinant human granulocyte colony stimulating factor. Radiotherapy was administered after chemotherapy according to guidelines from the NCCN, St Gallen International Consensus, and Chinese Breast Cancer Society. Follow-up assessments included physical examinations, breast ultrasonography, and abdominal ultrasonography every three months during years 1 and 2, every six months during years 3-5, and annually thereafter. Mammography and computed tomography of the chest were done annually. Outcomes The primary endpoint was disease-free survival, defined as the time from randomisation to the occurrence of a first local or regional or distant event, contralateral or second primary tumours, or death from any cause. Patients still alive without an event of interest were censored at the date of the last visit. Secondary endpoints included recurrence-free survival, distant disease-free survival, overall survival, and safety. We defined recurrence-free survival as the time from randomisation to the occurrence of invasive local, regional, and distant recurrences and death from any cause; distant disease-free survival as the time to the first distant recurrence or death from any cause; and overall survival as the time from randomisation to the date of death from any cause. All events were measured from the date of random assignment. Safety assessments included evaluations of the incidence and severity of adverse events occurring during chemotherapy or within 30 days of completing it, according to the National Cancer Institute Common Toxicity Criteria version 4.0. Statistical analysis We assumed that the percentage of patients who would be disease-free at three years was approximately 83% in the control arm and 89% in the carboplatin arm. A total of 144 disease-free survival events would provide 80% power to detect the difference between the two arms at a two sided significance level of 5%, on the basis of a 1:1 randomisation ratio. The planned enrolment period was 36 months, with an additional 36 months of follow-up after enrolment of the last patient. Assuming an annual dropout rate of 5%, a total of 808 patients would be needed. Given the possibility that the event incidence may be lower than anticipated, the primary analysis will be done when the cumulative number of events reaches 144 or three years after the last patient is enrolled, whichever comes first. We used the intention-to-treat population efficacy analyses. Safety and toxicity were assessed in patients receiving at least one protocol therapy dose. We estimated survival outcomes by using the Kaplan-Meier method. We used an unadjusted Cox proportional hazards model to estimate hazard ratios and confidence intervals, and we tested the comparison of survival between trial arms by log-rank testing. We assessed the assumptions of proportional hazards by assessment of the Schoenfeld residuals. When proportional hazards assumption was not valid, we provided restricted mean survival analysis and the piecewise hazard ratio within pre-defined time periods (0-12, 12-36, and >36 months). We also showed hazard ratios over time derived from the natural spline (df=4) smooth of scaled Schoenfeld residuals on the original timescale to provide an intuitive depiction of how the treatment effect evolved. We also provided adjusted hazard ratios for the primary endpoint of disease-free survival adjusted for menopausal status, pathological N stage, pathological T stage, histological grade, lymphovascular invasion, BRCA1/2 mutation status, homologous recombination repair related gene alterations, and Ki-67, which we modelled as a continuous variables using splines to capture potential non-linear effects. We did pre-specified subgroup analyses using a Cox proportional hazards model that included trial arms, subgroup factor, and the trial-group-by-subgroup interaction term for categorical variables. We also reported corresponding interaction hazard ratios and their 95% confidence intervals. To explore potential non-linear and interaction effects between Ki-67 expression and treatment effects on disease-free survival, we applied a restricted cubic spline with four knots. P values were two sided and considered significant at an overall significance level of 0.05. As no formal type I error control was implemented, all secondary endpoint analyses should be interpreted as exploratory. We used SPSS version 22.0 and R 4.4.1 for statistical analyses. Patient and public involvement Patient and public involvement is increasingly recognised as being important, but this research was initiated before formal patient and public involvement procedures became common in China. In addition, to maintain the confidentiality of clinical data, the patients were not involved in data analysis, interpretation, or writing up of the results. Although patients and the public were not directly involved in this trial, mainly owing to training restrictions, we informed patients about the trial and ensured their awareness of its purpose and content during recruitment. We also invited a member of the public to review our manuscript after submission, and the results were communicated to patients who expressed an interest during clinic visits. Results Patients’ characteristics Between March 2020 and March 2022, 808 patients were enrolled in China and randomly assigned to two treatment arms: 404 to the control arm and 404 to the carboplatin arm (fig 1). The baseline characteristics were well balanced between the arms (table 1). The median age was 47 (range 22-70) years at the time of study entry. A total of 441 (54.6%) patients had a primary tumour of >2 cm, and 296 (36.6%) patients had positive lymph nodes. Most patients (381/404; 94.3%) in the control arm received dose dense anthracyclines. A total of 398 (98.5%) patients in the control arm and 380 (94.1%) patients in the carboplatin arm completed chemotherapy. Dose reductions were needed for 89 (22.0%) patients in the control arm and 105 (26.1%) patients in the carboplatin arm. Cycle delays occurred in 18.8% (76/404) patients in the control arm and 26.3% (106/403) patients in the carboplatin arm. Haematological toxicity was the most frequent reason for dose reduction and cycle delay. A total of 23 (2.8%) patients (13 (3.2%) in the control arm and 10 (2.5%) in the carboplatin arm) experienced treatment delays attributable to covid-19. Flowchart of study. *Unspecified reasons Table 1Characteristics of patients in intention-to-treat population at baseline. Values are numbers (percentages) unless stated otherwiseCharacteristicsControl (n=404)Carboplatin (n=404)Median (range) age, years48 (24-70)47 (22-70)Menopausal status: Premenopausal240 (59.4)251 (62.1) Postmenopausal164 (40.6)153 (37.9)Histological grade: I-II38 (9.4)41 (10.1) III366 (90.6)363 (89.9)Pathological T stage: pT1183 (45.3)184 (45.5) pT2209 (51.7)206 (51.0) pT312 (3.0)14 (3.5)Pathological N stage: pN0255 (63.1)257 (63.6) pN199 (24.5)109 (27.0) pN238 (9.4)28 (6.9) pN312 (3.0)10 (2.5)Pathological tumour stage: I132 (32.7)125 (30.9) II218 (54.0)237 (58.7) III54 (13.4)42 (10.4)Lymphovascular invasion: Negative271 (67.1)273 (67.6) Positive133 (32.9)131 (32.4)Ki-67 index: <50%48 (11.9)45 (11.1) ≥50%356 (88.1)359 (88.9)EC schedule: Every 2 weeks381 (94.3)404 (100.0) Every 3 weeks23 (5.7)0 (0.0)BRCA1/2 genes: Deleterious variant30 (7.4)19 (4.7) No deleterious variant70 (17.3)71 (17.6) Unknown304 (75.2)314 (77.7)HRR related genes Deleterious variant21 (5.2)20 (5.0) No deleterious variant79 (19.6)70 (17.3) Unknown304 (75.2)314 (77.7)EC=epirubicin and cyclophosphamide; HRR=homologous recombination repair. Characteristics of patients in intention-to-treat population at baseline. Values are numbers (percentages) unless stated otherwise EC=epirubicin and cyclophosphamide; HRR=homologous recombination repair. Primary efficacy outcome As of the data cut-off date of 10 March 2025, the median follow-up time was 44.7 months. A total of 104 (12.9%) disease-free survival events occurred, including 79 local, regional, or distant relapses, 18 second primary malignancies, and seven contralateral breast cancers, as summarised in table 2. Estimated percentages of patients who would be disease-free at three years were 92.3% in the carboplatin arm and 85.8% in the control arm, with a statistically significant difference (unadjusted hazard ratio 0.64, 95% confidence interval (CI) 0.43 to 0.95; P=0.03) (fig 2, top). An additional Cox regression analysis adjusting for pre-specified prognostic factors was consistent with the unadjusted analyses, yielding an adjusted hazard ratio of 0.65 (95% CI 0.44 to 0.96; P=0.03). We found the proportional hazards assumption not to be valid for disease-free survival (P=0.02). A piecewise Cox model showed that the treatment effect changed over time: the hazard ratio was 0.31 (95% CI 0.13 to 0.73) for 0-12 months, 0.65 (0.39 to 1.09) for 12-36 months, and 1.98 (0.69 to 5.69) for >36 months (fig 2, bottom; supplementary table A). The hazard ratio curve over time shown in figure 2 (bottom graph) also showed that the hazard ratio was not constant over time: a pronounced early benefit was apparent within the first 12 months (point estimate well below 1 with confidence interval largely below 1), followed by attenuation between 12 and 36 months (point estimate below 1 but confidence interval overlapping 1), with the point estimate rising above 1 with wide uncertainty beyond 36 months. We also did restricted mean analyses that do not rely on the proportional hazards assumption, including the restricted mean survival time and restricted mean time lost. The difference in restricted mean survival time was 2.4 (95% CI 0.7 to 4.1; P=0.006) months, and the restricted mean time lost ratio was 0.59 (95% CI 0.40 to 0.86; P=0.006), supporting the benefit of adding carboplatin (supplementary table B). Table 2First disease-free survival event by treatment. Values are numbers (percentages) unless stated otherwise EventsControl (n=404)Carboplatin (n=404)Follow-up time, months: Mean42.743.9 Median44.345.4 Minimum to maximum3.0-58.89.1-57.3Any event62 (15.3)42 (10.4)Local and regional recurrence6 (1.5)3 (0.7)Distant metastasis:43 (10.6)27 (6.7) Lung25 (6.2)16 (4.0) Liver15 (3.7)7 (1.7) Bone19 (4.7)13 (3.2) Brain2 (0.5)7 (1.7) Other22 (5.4)13 (3.2)Second primary malignancy10 (2.5)8 (2.0)Contralateral breast cancer3 (0.7)4 (1.0)Death0 (0)0 (0) First disease-free survival event by treatment. Values are numbers (percentages) unless stated otherwise Primary efficacy outcome. Top: Kaplan-Meier plot showing disease-free survival for patients. Bottom: non-linear plot showing hazard ratio (HR) over time derived from natural spline (df=4) smooth of scaled Schoenfeld residuals on original timescale with y axis relabelled in HR units. CI=confidence interval Secondary efficacy outcomes For secondary endpoints, treatment with carboplatin was associated with improvement in three year recurrence-free survival (93.8%) compared with control (88.3%; hazard ratio 0.59, 95% CI 0.37 to 0.93; P=0.02) (fig 3, top), in three year distant disease-free survival (94.8% v 89.8%; hazard ratio 0.61, 0.37 to 0.98; P=0.04) (fig 3, middle), and in three year overall survival (98.0% v 94.0%; hazard ratio 0.41, 0.20 to 0.83; P=0.01) (fig 3, bottom). Secondary efficacy outcomes. Kaplan-Meier plots showing recurrence-free survival (top panel), distant disease-free survival (middle panel), and overall survival (bottom panel). CI=confidence interval Exploratory subgroup analysis Figure 4 shows the exploratory subgroup analyses for disease-free survival; the interaction test did not indicate significant heterogeneity, suggesting that the treatment effect was consistent across all subgroups, including menopausal status, pathological N stage, pathological T stage, pathological tumour stage, histological grade, lymphovascular invasion, BRCA1/2 mutation status, and homologous recombination repair related gene alterations. For stage I disease, the estimated percentage of patients who would be disease-free at three years was 92.7% in the carboplatin arm and 90.1% in the control arm; for stage II-III disease, the estimated percentage of patients who would be disease-free at three years was 92.1% in the carboplatin arm and 83.7% in the control arm (supplementary figure A). The formal test of interaction between treatment and continuous Ki-67 also yielded a non-significant result (interaction P=0.75), indicating no evidence that treatment effect varied across the spectrum of Ki-67 expression (supplementary figure B). Analysis by nodal status suggested that Ki-67 does not seem to be a treatment effect modifier in either node negative (interaction P=0.73) or node positive patients (interaction P=0.29) (supplementary figure C). Exploratory subgroup analyses for disease-free survival. An interactive version of this graphic and downloadable data are available at https://public.flourish.studio/visualisation/26214405/ Safety All 807 patients in the safety population experienced at least one treatment related adverse event (table 3). The overall incidence of grade 3 or 4 treatment related adverse events was 222 (55.0%) of 404 patients in the control arm and 269 (66.7%) of 403 patients in the carboplatin arm. The most common treatment related adverse events of grade 3 or 4 were haematological toxicities; grade 3 or 4 events occurring more frequently in the carboplatin arm (percentage point difference ≥2% versus control arm) included neutropenia, thrombocytopenia, anaemia, and infusion related reactions. Overall, 30 treatment related serious adverse events occurred: 13 (3.2%) in the control arm (febrile neutropenia, n=7; infection, n=3; diarrhoea, vomiting, and liver dysfunction, n=1 each) and 17 (4.2%) in the carboplatin arm (febrile neutropenia, n=6; infection, n=5; severe anaemia and thrombocytopenia, n=2 each; rash and thrombosis, n=1 each). Discontinuation of study treatment due to toxicity was recorded in 2 (0.5%) and 18 (4.5%) patients, respectively. We found no significant difference in the incidence of treatment related serious adverse events or treatment related adverse events leading to dose discontinuation (supplementary table C). No treatment related deaths occurred in the study. Table 3Most common treatment related adverse events according to treatment arms. Values are numbers (percentages) unless stated otherwiseAdverse events*Control (n=404)Carboplatin (n=403)All gradesGrade 3-4All gradesGrade 3-4Any404 (100.0)222 (55.0)403 (100.0)269 (66.7) Haematological Neutropenia376 (93.1)216 (53.5)397 (98.5)254 (63.0)Anaemia317 (78.5)7 (1.7)341 (84.6)18 (4.5)Thrombocytopenia188 (46.5)14 (3.5)243 (60.3)28 (6.9)Febrile neutropenia20 (5.0)20 (5.0)19 (4.7)19 (4.7) Non-haematological Fatigue287 (71.0)19 (4.7)306 (75.9)25 (6.2)Myalgia262 (64.9)4 (1.0)271 (67.2)3 (0.7)SGPT increased236 (58.4)11 (2.7)224 (55.6)14 (3.5)Nausea215 (53.2)12 (3.0)206 (51.1)14 (3.5)SGOT increased173 (42.8)6 (1.5)165 (40.9)3 (0.7)Peripheral sensory neuropathy148 (36.6)17 (4.2)133 (33.0)20 (5.0)Insomnia116 (28.7)3 (0.7)119 (29.5)1 (0.2)Vomiting97 (24.0)6 (1.5)102 (25.3)8 (2.0)Constipation47 (11.6)058 (14.4)0Diarrhoea37 (9.2)1 (0.2)34 (8.4)1 (0.2)Palpitations28 (6.9)042 (10.4)0Infection27 (6.7)3 (0.7)31 (7.7)1 (0.2)Pain (other than musculoskeletal)20 (5.0)020 (5.0)0Stomatitis18 (4.5)029 (7.2)0Rash18 (4.5)025 (6.2)1 (0.2)Syncope3 (0.7)3 (0.7)4 (1.0)4 (1.0)Thrombosis3 (0.7)01 (0.2)1 (0.2)Infusion related reaction1 (0.2)017 (4.2)7 (1.7)SGOT=serum glutamic oxaloacetic transaminase; SGPT=serum glutamic pyruvic transaminase.*Treatment related adverse events of grade 1-2 occurring in ≥5% of patients and all events of grade ≥3 occurring in treatment arms are listed. No grade 5 treatment related adverse events occurred. Most common treatment related adverse events according to treatment arms. Values are numbers (percentages) unless stated otherwise SGOT=serum glutamic oxaloacetic transaminase; SGPT=serum glutamic pyruvic transaminase. Treatment related adverse events of grade 1-2 occurring in ≥5% of patients and all events of grade ≥3 occurring in treatment arms are listed. No grade 5 treatment related adverse events occurred. Discussion The CITRINE trial was designed to explore, for the first time, whether the addition of carboplatin to an anthracycline/taxane based adjuvant chemotherapy regimen could improve the prognosis of patients with high risk, early stage, triple negative breast cancer. The results indicated the superior benefits of the four drug regimen including carboplatin compared with the standard anthracycline/taxane regimen, with a statistically significant improvement in disease-free survival (92.3% v 85.8%; hazard ratio 0.64) and overall survival without new safety concerns. Key features of this study include that the target population is clearly delineated, focusing specifically on patients at high risk who need adjuvant chemotherapy, as opposed to a combination of neoadjuvant and adjuvant populations, and that patients in the control arm uniformly received sequential weekly paclitaxel after anthracyclines, representing the optimised schedule recommended by ECOG E1199, with 94.3% of cases adhering to the dose dense regimen. Subgroup analyses of disease-free survival indicated that outcomes consistently favoured the carboplatin arm in pre-specified subgroups. Comparison with other studies In adjuvant therapy, the optimal treatment regimen for triple negative breast cancer remains to be further elucidated.27 Beyond anthracycline/taxane based regimens, the intensification of adjuvant treatment for triple negative breast cancer represents an area of significant research interest, exploring strategies such as poly-ADP ribose polymerase (PARP) inhibitors, immunotherapy, and capecitabine. The OlympiA study showed that the PARP inhibitor olaparib as a one year intensified adjuvant treatment can significantly improve the invasive disease-free survival and overall survival of patients at high risk carrying BRCA1/2 germline mutations.28 29 However, the prevalence of BRCA1/2 variants in unselected patients with triple negative breast cancer is low, at approximately 11%,30 limiting the broader application of PARP inhibitors. Several trials have shown the efficacy of immunotherapy in the neoadjuvant setting of triple negative breast cancer, whereas the value of immunotherapy in the adjuvant treatment of triple negative breast cancer remains controversial.31 32 33 34 The IMpassion030 study failed to confirm that the addition of atezolizumab to weekly paclitaxel followed by an anthracycline regimen could improve the prognosis of patients with triple negative breast cancer.35 Conversely, the A-BRAVE trial showed that although one year of adjuvant avelumab did not significantly improve disease-free survival in patients with high risk triple negative breast cancer, the secondary endpoint of overall survival was markedly improved compared with control.36 Regarding capecitabine application, the SYSUCC-001 study showed that using low dose metronomic capecitabine maintenance therapy after standard chemotherapy improved disease-free survival in patients with triple negative breast cancer but had no significant effect on overall survival.37 These data highlight the urgent need for more effective treatment strategies to optimise adjuvant therapy for the broad population of patients with triple negative breast cancer. Platinum based chemotherapy regimens have shown consistent efficacy in the neoadjuvant treatment of early triple negative breast cancer and treatment of advanced triple negative breast cancer,14 22 although their value as adjuvant treatment remains debatable. The incorporation of platinum based drugs into neoadjuvant chemotherapy regimens has been shown to increase the pathological complete response rate and may further improve patients’ survival outcomes, although some study results remain inconsistent.15 16 17 18 19 20 21 38 39 On the basis of these findings, the KEYNOTE-522, CamRelief, and NeoTRIP trials established a neoadjuvant chemotherapy backbone consisting of carboplatin as standard therapy.31 32 33 In the context of adjuvant therapy, the PATTERN study showed that a non-anthracycline regimen of six cycles of weekly paclitaxel plus carboplatin improved disease-free survival compared with three cycles of cyclophosphamide, epirubicin, and fluorouracil followed by three cycles of docetaxel, although no significant effect on overall survival was observed.40 Thus, for patients who cannot receive anthracyclines, such as older patients and those with cardiac disease, an adjuvant taxane/carboplatin doublet might be a reasonable alternative. The PEARLY trial indicated the survival benefits of incorporating carboplatin into an eight cycle regimen of anthracycline followed by taxane in both neoadjuvant and adjuvant settings, although only 29.4% of patients received adjuvant therapy.41 42 Moreover, not all patients received the weekly paclitaxel regimen; rather, a subset of patients were treated with docetaxel on a three weekly schedule. Therefore, whether inclusion of carboplatin in the adjuvant setting of triple negative breast cancer could further improve survival remains unclear, particularly in the context of the optimal regimen involving sequential weekly paclitaxel following dose dense anthracyclines. The CITRINE study innovatively bridged this gap by showing for the first time that the combination of weekly paclitaxel and carboplatin, following dose dense anthracyclines, significantly improves survival compared with the traditional regimen, with a 6.5% absolute improvement in three year disease-free survival. Our analysis showed that the proportional hazards assumption for disease-free survival did not hold, and the hazard ratio was non-constant over time. Specifically, piecewise Cox models across the three pre-specified periods (0-12, 12-36, and >36 months) and the plot of hazard ratio over time indicated that the benefit was concentrated in the first year, diminished through 12 to 36 months, and was highly uncertain beyond 36 months. The number of disease-free survival events occurring beyond 36 months was relatively limited, with a total of 11 disease-free survival events in the carboplatin arm, of which six were classified as second primary or contralateral breast cancers, compared with five disease-free survival events in the control arm, of which three were also second primary or contralateral breast cancers. This makes the hazard ratio in this late window sensitive to single events, necessitating cautious interpretation and highlighting the need for extended follow-up to fully assess the durability of benefit. Nevertheless, the time pattern informs clinicians that the addition of carboplatin primarily reduces early recurrence risk, whereas longer term differences become less certain. The NRG-BR003 study, which shares a similar design with this study, was presented at the 2025 American Society of Clinical Oncology meeting.43 Although the study did not achieve statistical significance, the 4.9% absolute improvement in five year invasive disease-free survival was considered clinically meaningful. Moreover, notable differences between NRG-BR003 and our study exist in terms of patient population, the time span of the study endpoint, and the definition of high risk disease. Furthermore, variations in race and genetic background may represent one of the potential contributing factors. Moreover, exploratory stage stratified analyses showed a numerical trend towards greater absolute benefit in stage II/III disease than in stage I disease (hazard ratio 0.61 for stage II/III versus 0.76 for stage I), but the test for interaction between stage (I versus II/III) and carboplatin effect was not significant (P for interaction=0.63). This indicated that the effect of carboplatin on disease-free survival showed the same clinical benefits in stage I and stage II/III. Our trial did not enrol “pure” stage I patients, but rather those with stage I triple negative breast cancer with high risk biology—specifically, node negative tumours with Ki-67 ≥50%. These patients have a relatively high relapse risk comparable to that seen in stage II/III disease, justifying the use of an intensified regimen.44 45 As for stage I disease with low risk, treatment de-escalation should be explored in the future. For example, paclitaxel combined with carboplatin without anthracycline, which has been confirmed by both the PATTERN study and the WSG-TN study, is an optional treatment for these patients.40 46 Furthermore, in selected subpopulations of patients with triple negative breast cancer, investigations are underway to assess the feasibility of completely omitting chemotherapy (for example, in the low risk population with high tumour infiltrating lymphocytes). The endeavour to determine whether to integrate or omit treatment modalities transcends the confines of binary reasoning, unfolding instead as a richly layered and intricately woven pursuit. The spectrum of adverse events was consistent with those reported for dose dense anthracyclines,8 weekly paclitaxel,5 40 and carboplatin,21 40 with no new safety concerns identified. As expected, a higher incidence of grade 3 or 4 treatment related adverse events was seen among patients in the carboplatin arm compared with counterparts in the control arm (66.7% v 55.0%), mostly driven by haematological toxicities including neutropenia, anaemia, and thrombocytopenia. The incidence of febrile neutropenia did not differ significantly between the two arms and predominantly occurred during the dose dense anthracycline regimen. All patients receiving dose dense anthracycline were advised to receive prophylactic management with pegylated recombinant human granulocyte colony stimulating factor, which decreased the risk of infection, shown by the presence of febrile neutropenia.47 48 49 Compared with the control arm, both all grade and grade 3-4 anaemia/thrombocytopenia were more frequently observed in the carboplatin arm. These adverse events were generally manageable through conventional supportive care or dose modifications. Only one patient in the carboplatin arm needed transfusion as a result of anaemia. Furthermore, infusion related reactions attributed to carboplatin warrant careful attention, as they resulted in the discontinuation of carboplatin treatment in 13 patients. These reactions predominantly manifested during the final cycle of chemotherapy. Notably, the incorporation of carboplatin did not lead to an increased incidence of treatment related serious adverse events for the carboplatin arm versus control (4.2% v 3.2%). Additionally, a comparable proportion of patients in both arms completed the full cycles per protocol (94.3% v 98.5%). Thus, the generally manageable safety profile with carboplatin containing regimens facilitated the maximisation of exposure to treatment to achieve favourable long term cancer related outcomes in a potentially curable disease setting. Limitations of study A possible limitation of this study pertains to the fact that he KEYNOTE-522 trial has shown considerable clinical advantages, leading to growing recognition of the combination of chemotherapy and immunotherapy as the favoured therapeutic approach for patients with stage II–III triple negative breast cancer.50 Instead of opting for immediate surgery followed by adjuvant chemotherapy, recent evidence advocates for initiating neoadjuvant therapy that includes pembrolizumab. Consequently, the design of our study deviates to some degree from the current standard of care. However, several contextual factors should be taken into account. The KEYNOTE-522 trial is undoubtedly a landmark study with the potential to reshape clinical practice. Nevertheless, it holds an NCCN category 2A recommendation, which indicates expert consensus based on lower level evidence. Moreover, the survival benefit associated with combining neoadjuvant immunotherapy and chemotherapy remains largely dependent on data exclusively derived from the KEYNOTE-522 trial. We also note that, at the time this study was initiated, pembrolizumab had not yet been granted approval in the neoadjuvant setting. A lag period typically occurs before new adjuvant therapies are widely adopted in clinical practice. Moreover, preoperative clinical staging may occasionally underestimate pathological stage, prompting a surgery first approach. As a result, a subgroup of patients who cannot be definitively classified as stage II–III until after surgical resection and pathological evaluation inevitably exists. Thus, our study provides a valuable adjuvant option for resectable triple negative breast cancer with high risk features, but it neither replaces neoadjuvant chemotherapy combined with immunotherapy for locally advanced disease nor invalidates the multidimensional benefits of preoperative therapy. As management of triple negative breast cancer evolves, our goal remains to match tumour biology with context specific strategies, and treatment selection should be guided by clinical context and risk stratification. This study has several other limitations that warrant consideration. Firstly, concerns about the tolerability of the dose dense regimen precluded setting the control arm uniformly to dose dense anthracyclines followed by weekly paclitaxel, thereby introducing some heterogeneity and bias into the control arm. However, on primary completion of the study, we observed that 94.3% of the control arm participants received the same two weekly anthracycline regimen as those in the carboplatin arm. Secondly, all participants were Chinese, raising concerns about racial variability and requiring additional evidence to confirm the generalisability of these conclusions to other populations. Thirdly, the open label design of the study might introduce potential biases affecting the objectivity of the results. However, all outcome assessors and data collectors were masked to treatment assignment to minimise the potential for bias. Fourthly, more than 75% of the participants were not tested for BRCA gene status. The absence of this data may hinder further investigation into the association between the efficacy of the platinum agent and BRCA status. Lastly, long term survival data remain immature, highlighting the need for extended follow-up to fully assess the durability of the treatment’s benefits. Particularly with regard to overall survival, the current follow-up duration is insufficient to draw definitive conclusions. Conclusions This study found that the addition of carboplatin to standard adjuvant anthracycline/taxane based chemotherapy significantly improved survival outcomes in patients with high risk, early stage, triple negative breast cancer, driven by reduction of early recurrence risk. Despite an increased incidence of adverse events, primarily haematological, the safety profile of carboplatin containing chemotherapy was consistent with the known safety profiles of the relevant drugs. What is already known on this topic The prognosis of high risk, early stage, triple negative breast cancer is unsatisfactory, and a need remains to optimise adjuvant intensification strategies The prognosis of high risk, early stage, triple negative breast cancer is unsatisfactory, and a need remains to optimise adjuvant intensification strategies The benefit of incorporating carboplatin into adjuvant anthracycline/taxane based chemotherapy in triple negative breast cancer remains controversial The benefit of incorporating carboplatin into adjuvant anthracycline/taxane based chemotherapy in triple negative breast cancer remains controversial What this study adds This study is the first to add carboplatin to anthracycline/taxane based adjuvant chemotherapy regimen for patients with high risk, early stage, triple negative breast cancer This study is the first to add carboplatin to anthracycline/taxane based adjuvant chemotherapy regimen for patients with high risk, early stage, triple negative breast cancer Carboplatin intensified chemotherapy significantly improved disease-free survival, primarily by reducing early recurrence risk, as well as overall survival, compared with standard chemotherapy Carboplatin intensified chemotherapy significantly improved disease-free survival, primarily by reducing early recurrence risk, as well as overall survival, compared with standard chemotherapy Addition of carboplatin into adjuvant anthracycline/taxane based chemotherapy raised no new safety concerns Addition of carboplatin into adjuvant anthracycline/taxane based chemotherapy raised no new safety concerns Acknowledgments We are grateful to the patients and their families who contributed to this study. Extra material supplied by authors Web appendix: Supplementary tables and figures Web appendix: Study protocol Web appendix: Analytical code Contributors: YL, YG, XZZ, GYL and KDY contributed equally (joint first authors). LF, ZHW, and ZMS contributed equally (joint last authors). YL, YG, XZZ, GYL, KDY, LF, ZHW, and ZMS contributed to the study design. YL, YG, XZZ, GYL, KDY, FY, LC, MH, ZH, CMC, AYC, JJL, YFH, GHD, JW, YZJ, LF, ZHW, and ZMS contributed to the collection and assembly of data. YL, YG, XZZ, GYL, KDY, FY, LC, MH, LF, ZHW, and ZMS contributed to the analysis and interpretation of data. YL, YG, XZZ, and FY did the statistical analysis. YL, YG, XZZ, GYL, KDY, FY, LF, ZHW, and ZMS drafted the manuscript. All authors contributed to revision of the manuscript and approved the final draft. LF, ZHW, and ZMS are the guarantors. The corresponding author attests that all listed authors meet authorship criteria and that no others meeting the criteria have been omitted. Funding: This work was supported by grants from the National Key Research and Development Project of China (2020YFA0112304, 2023YFF1205004), the National Natural Science Foundation of China (82403912, 91959207, 92159301, 82473339, 82472866, and 82103369), the Shanghai Anti-Cancer Association (SACA-AX202314), the Science and Technology Commission of Shanghai Municipality (22Y11912800), the Shanghai Key Laboratory of Breast Cancer (12DZ2260100), the SHDC Municipal Project for Developing Emerging and Frontier Technology in Shanghai Hospitals (SHDC12021103), and the China Postdoctoral Science Foundation (2022M710757). The funders had no role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript. Competing interests: All authors have completed the ICMJE uniform disclosure form at https://www.icmje.org/disclosure-of-interest/ and declare: support from the National Key Research and Development Project of China, the National Natural Science Foundation of China, the Shanghai Anti-Cancer Association, the Science and Technology Commission of Shanghai Municipality, the Shanghai Key Laboratory of Breast Cancer, the SHDC Municipal Project for Developing Emerging and Frontier Technology in Shanghai Hospitals, and the China Postdoctoral Science Foundation for the submitted work; no financial relationships with any organisation that might have an interest in the submitted work in the previous three years; no other relationships or activities that could appear to have influenced the submitted work. Transparency: The lead authors (the manuscript’s guarantors) affirm that the manuscript is an honest, accurate, and transparent account of the study being reported; that no important aspects of the study have been omitted; and that any discrepancies from the study as planned and registered have been explained. Dissemination to participants and related patient and public communities: The results can be communicated to study participants who express an interest during clinic visits. The findings will also be internally disseminated through the media departments and websites of the participating institutes, shared with public health organisations, and presented at regional and national events with health, lay, and government representation. Provenance and peer review: Not commissioned; externally peer reviewed. Ethics statements Ethical approval This study was approved by the independent ethics committee of Fudan University Shanghai Cancer Center. All participants gave written informed consent. Data availability statement The code used to analyse the data in the paper can be found in the supplementary materials. The data underlying the findings in this paper are openly and publicly available and can be found at Fudan University Shanghai Cancer Center Clinical Trial Database (http://117.72.180.52:8081 [search for CT2025-00002]). If you encounter problems accessing the data, please contact the corresponding author. References Associated Data Supplementary Materials Web appendix: Supplementary tables and figures Web appendix: Study protocol Web appendix: Analytical code Data Availability Statement The code used to analyse the data in the paper can be found in the supplementary materials. The data underlying the findings in this paper are openly and publicly available and can be found at Fudan University Shanghai Cancer Center Clinical Trial Database (http://117.72.180.52:8081 [search for CT2025-00002]). If you encounter problems accessing the data, please contact the corresponding author.