Summit and Akeso Land HARMONi in The Lancet Oncology as Ivonescimab Beats Durvalumab on Survival in a China Biliary Tract Cancer Trial

Eleven months after the topline, the confidence intervals arrive

Summit Therapeutics disclosed the HARMONi topline in September 2025 and has been repeating the headline numbers ever since. The peer-reviewed manuscript, published August 25, 2026 and authored by Xiuning Le of The University of Texas MD Anderson Cancer Center with Li Zhang of Sun Yat-sen University Cancer Center as senior author, fills in the design and the denominators. Patients had stage IIIB to IV non-squamous EGFR-mutated NSCLC, ECOG 0 or 1, and progression after a third-generation EGFR TKI. They received ivonescimab 20 mg/kg or placebo with pemetrexed 500 mg/m2 and carboplatin AUC 5 every three weeks, stratified by brain metastasis status and region. Asian sites capped enrollment at age 75.

The PFS analysis, assessed by blinded independent radiology review, counted 275 progression or death events: 129 among 172 patients in the ivonescimab arm and 146 among 173 in the placebo arm, according to the abstract. Median PFS was 6.8 months (95% CI 5.7-7.1) against 4.4 months (4.1-5.5). Summit's release says the benefit held across preplanned subgroups.

Overall survival is where the paper will be read most closely. At a median follow-up of 29.7 months there were 122 deaths in the ivonescimab group and 140 in the placebo group, for medians of 16.8 months (14.3-19.0) and 14.0 months (12.8-15.7). The hazard ratio of 0.79 carries a 95% confidence interval of 0.62 to 1.01, so the primary OS endpoint was not met. Summit has since reported a June 2026 data cut, announced July 22, in which western patients had an OS hazard ratio of 0.76, and it will present the updated analysis at the IASLC World Conference on Lung Cancer in a 10-minute slot on September 15, 2026.

Safety in the published dataset

Grade 3-4 treatment-related adverse events were led by cytopenias in both arms. Decreased neutrophil count occurred in 42 of 218 patients (19%) on ivonescimab versus 36 of 218 (17%) on placebo, decreased white cell count in 13% versus 11%, decreased platelets in 12% versus 6%, and anemia in 10% versus 12%. Serious treatment-related events were nearly twice as common with the bispecific, 61 patients (28%) versus 33 (15%). Four patients in the ivonescimab arm died of treatment-related causes, one each from disease progression, multiple organ dysfunction and hepatic failure, and one from gastrointestinal hemorrhage with pulmonary embolism. Five placebo patients died of treatment-related pneumonitis, myocardial infarction, cerebrovascular accident, cognitive disorder and embolic stroke. Summit's release adds that grade 3-5 treatment-related hemorrhage, the VEGF-class concern, stayed under 1%.

"Once a patient with EGFR-mutated lung cancer progresses after a third-generation EGFR TKI, there are limited treatment options for those patients and the survival may be limited," Le said in the Summit release. "In this patient population, traditional anti-PD-(L)1 therapies have not established a clear benefit in prior Phase III studies."

HARMONi-GI1 and the durvalumab comparator

Akeso's announcement, timed at 5 p.m. Eastern on August 25, reports that an Independent Data Monitoring Committee reviewed a pre-specified interim analysis of AK112-309, also called HARMONi-GI1, and found the study had met its overall survival primary endpoint. Progression-free survival and objective response rate, the key secondary endpoints, were also met. The trial's ClinicalTrials.gov record lists 682 patients enrolled since October 20, 2024, with gemcitabine and cisplatin as the chemotherapy backbone in both arms. Durvalumab plus that regimen has been the guideline first-line immunotherapy for advanced biliary tract cancer since the TOPAZ-1 trial, and Akeso describes it as "the global gold standard."

The company calls HARMONi-GI1 "the first Phase III study in biliary tract cancer to demonstrate a statistically significant positive OS result" against a PD-L1 antibody plus chemotherapy. That claim, and the outcome itself, rest for now on a press release. No hazard ratio, median, confidence interval or safety table has been published, and Akeso says the detail will come at "an upcoming international academic conference" and in a peer-reviewed journal. Summit's own statement is careful to note that the study is single-region, conducted in China, and that "all relevant data" were "exclusively generated, managed, and analyzed by Akeso."

Michelle Xia, Akeso's founder, chairwoman, president and CEO, framed it as a program milestone. "The positive result of HARMONi-GI1 marks the first positive Phase III study for ivonescimab in gastrointestinal tumors, following four positive Phase III results in lung cancer," she said.

Where the program stands

By Summit's count there are 16 Phase III ivonescimab studies announced, ongoing or completed: five Summit-sponsored global trials, one GORTEC cooperative-group study in head and neck cancer, and 10 Akeso trials in China. Summit's own list includes HARMONi-3 against pembrolizumab plus chemotherapy in first-line NSCLC, HARMONi-7 monotherapy against pembrolizumab in PD-L1-high disease, HARMONi-GI3 against bevacizumab plus chemotherapy in colorectal cancer, and HARMONi-GU1, a Phase II/III urothelial study pairing the bispecific with enfortumab vedotin. The drug has been approved in China since May 2024 and remains investigational in the United States and Europe. The FDA accepted the EGFR-mutated NSCLC BLA in January 2026.

"We continue to evaluate longer-term results from HARMONi as the data mature," said Maky Zanganeh, Summit's president and co-CEO, in the publication release.

Why This Matters to the APO|APE Reader

The FDA's November 14 decision will turn on a PFS benefit of 2.4 months and an OS hazard ratio whose upper bound sits at 1.01, and the agency's Oncology Center of Excellence has spent the past two years pressing sponsors on exactly that pattern, an immunotherapy combination with a clean PFS win and an unresolved survival signal. The 70% Asian enrollment in HARMONi is the other question a reviewer will ask, which is why Summit is leading with the western-subgroup OS hazard ratio of 0.76 ahead of WCLC. For pathology and molecular labs the near-term consequence is procedural: if the BLA is approved, the eligible population is defined by an EGFR mutation already documented at first-line testing plus progression on osimertinib or another third-generation TKI, with no new biomarker gate, so the demand shift is in re-biopsy and resistance-mechanism testing at progression, not in a companion diagnostic. The biliary result, if it holds up in a peer-reviewed table, would be the first time a bispecific has displaced a PD-L1 antibody in a gastrointestinal first-line setting, but a China-only trial against durvalumab will need a multiregional follow-up before it changes practice outside Akeso's home market.