HARMONi-2: ivonescimab extends overall survival over pembrolizumab in China

One bispecific antibody against pembrolizumab alone, no chemotherapy in either arm

Ivonescimab is a bispecific antibody that binds PD-1 and VEGF; the abstract describes it as first in class. HARMONi-2 set it against pembrolizumab, a PD-1 inhibitor, in a head-to-head design with no chemotherapy in either arm.

Eligible patients had treatment-naïve, locally advanced or metastatic NSCLC with a PD-L1 tumour proportion score (TPS) of 1% or higher and no EGFR or ALK alterations. Between November 2022 and August 2023, 398 patients were randomised 1:1 to ivonescimab 20 mg/kg (n=198) or pembrolizumab at a 200 mg flat dose (n=200), given every three weeks. The abstract states that enrolment took place at centres in China, and all fifteen affiliations in the author block are in China. The trial has no sites outside the country.

The primary endpoint was progression-free survival (PFS) assessed by an independent radiology review committee under RECIST v1.1, and it was met in the earlier analysis: median PFS was 11.1 months with ivonescimab and 5.8 months with pembrolizumab (HR 0.51, 95% CI 0.38–0.69, P<0.0001). Overall survival (OS) was the key secondary endpoint, to be tested at two planned interim analyses and one final analysis. The WCLC presentation covers the second interim analysis.

Overall survival cleared its interim boundary at 234 deaths

The second interim analysis was planned for 232 OS events, with a one-sided alpha of 0.0139. By the data cut-off of 20 August 2026, 234 events had occurred, and the prespecified O’Brien-Fleming spending function set the alpha for this look at 0.0141.

Median OS was 30.75 months with ivonescimab and 22.57 months with pembrolizumab, a difference of just over eight months at the median. The hazard ratio was 0.73 (95% CI 0.57–0.95), and the reported P value of 0.009 falls inside the alpha allotted to the analysis. The authors describe the result as a statistically significant and clinically meaningful survival benefit; the second of those judgements is theirs.

Several figures a reader would want are not in the abstract. It gives no median follow-up, no landmark survival rates at fixed timepoints, and no information on what treatment patients in either arm received after progression. Because the analysis is an interim one, a final OS analysis remains to come.

Squamous histology and PD-L1 TPS of 50% or higher show the largest effect

The abstract reports OS hazard ratios for four prespecified subgroups and calls the benefit generally consistent across them. In squamous histology the hazard ratio was 0.65 (95% CI 0.45–0.95); in non-squamous histology it was 0.79 (95% CI 0.55–1.14). Split by PD-L1 expression, the hazard ratio was 0.85 (95% CI 0.61–1.18) at TPS 1–49% and 0.58 (95% CI 0.38–0.89) at TPS 50% or higher.

Two of the four intervals include 1, those for non-squamous disease and for TPS 1–49%, and the abstract gives no patient or event counts for any subgroup, so the width of each interval cannot be related to its size. The point estimates put the larger effect in squamous tumours and in tumours with high PD-L1 expression.

More treatment-related and serious adverse events with ivonescimab, similar discontinuation rates

Treatment-related adverse events (TRAEs) of any grade occurred in 93.4% of patients on ivonescimab and 84.9% on pembrolizumab. Serious TRAEs were reported in 29.9% versus 21.6%. TRAEs leading to permanent discontinuation were 4.1% with ivonescimab and 5.0% with pembrolizumab, and immune-related adverse events were 34.0% versus 33.7%.

The abstract characterises the two profiles as broadly comparable and states that no new safety signals were identified with ivonescimab relative to the established pembrolizumab profile. It does not break the events down by type or by grade 3 or higher, and it does not report treatment-related deaths, so the events behind the higher rate of serious TRAEs in the ivonescimab arm cannot be identified from it.

A single-country trial, and the claim the authors build on it

Every patient in HARMONi-2 was treated in China, and the abstract makes no claim about other populations. Four of the twenty authors list Akeso Biopharma, Zhongshan, as their affiliation. The conclusion of the abstract goes beyond the OS figures: the authors write that the findings support ivonescimab as a chemotherapy-free first-line treatment with the potential to redefine the standard of care and to serve as a cornerstone of treatment in this population. That is the authors’ framing.

In the IASLC news release accompanying the presentation, Prof. Zhou summarised the result in the same terms as the abstract.

“Ivonescimab significantly prolonged overall survival versus pembrolizumab in this patient population, with no new safety signals identified relative to the established pembrolizumab safety profile.”Prof. Caicun Zhou, Shanghai Pulmonary Hospital, Tongji University School of Medicine, Shanghai, China

What the abstract establishes is narrower than the conclusion drawn from it: in a Chinese population with PD-L1-positive, EGFR- and ALK-negative advanced NSCLC, ivonescimab monotherapy prolonged OS over pembrolizumab monotherapy at a prespecified interim analysis, at the cost of more serious treatment-related events and with no difference in discontinuations. The final OS analysis is still pending.

Written by the Lung Summit editorial team.