ASCO 2026 第二部分：Akeso的HARMONi-6预览

Rebecca Liang, Ph.D.+852 2123 2656rebecca.liang@bernsteinsg.comEllie Li+852 2123 2621ellie.li@bernsteinsg.com China Pharma and BiotechAkeso Inc Rating Market-Perform Price Target 9926.HK 130.00 HKD ASCO 2026 Part 2 - Akeso's HARMONi-6 preview We expect the ASCO 2026 HARMONi-6 plenary to trade almost entirely on the interim OSHR and how it lines up against plenary-level expectations. Our base case is that the HRskirts around the O’Brien-Fleming boundary: either a“technical win” in the 0.70–0.72 Close Date27 May 20269926.HK Close Price (HKD)114.40Price Target (HKD)130.00Upside/(Downside)14%52-Week Range179.00/73.10ASIAX1,978.51FYEDecDiv YieldNAMarket Cap (HKD) (M)105,379EV (HKD) (M)86,504 range within the OBF boundary (i.e. statistically significant) with modest but consistent curveseparation, fully justifying the plenary slot; or“directionally positive” 0.72-0.74whereit may exceed the boundary but still show a favorable trend and allows shot-on-goal forfinal OS to clear the boundary.We expect positive to measured reaction to a “technicalwin” (10-20% upside that may fade with profit-taking), and muted to mildly negativereaction to “directionally positive”.There is a caveat that Akeso has not issued anyheadline release about securing statistically significant OS HR, which was their pattern forstrongly positive data in the past - but this could be in keeping with the strict non-disclosure Separately, AK112’s Phase II ASCO 2026 readout in 1L non-MSI-H/dMMR mCRC (ivo+mFOLFOX6) adds an early but encouraging data point.ORR of ~71% and 6-monthPFS rates of ~85–95% compare favorably cross-trial with bevacizumab- and cetuximab-based 1L regimens. That said, the dataset is small, non-randomized, and limited to earlyORR and 6-month PFS rate. Though modestly boosting confidence that the PD-L1/VEGF Looking past ASCO, the story is about a tight 2H26–2027 NSCLC readout lineup. The2H26HARMONi-3 squamous final PFS readout(1L PD-L1 all-comer vs pembro+chemo) is thenext big global efficacy test after an interim that missed its strict PFS bar and left the trialinconclusive for now. The14 Nov 2026 HARMONi PDUFAin 2L EGFRm, TKI-pretreatednsq-NSCLC is the first true registration swing and will determine how fast ivonescimab canscale outside China.(Continued on the next page) Investment Implications DETAILS (cont.) Into 1H27, the HARMONi-3 non-sq PFS data then set up whether the drug can broaden its 1L reach beyond squamous.Overall, we still see ASCO 2026 as a checkpoint rather than the finish line: even with strong H-6 and supportive CRC data,investors will likely stay balanced until HARMONi-3 final PFS/OS are in hand. HARMONI-6 OS PREVIEW ASCO plenary sets a high bar; required interim OS HR in the range 0.72-0.74 for statistical significanceFollowing the previous reported positive PFS outcome, HARMONi-6 will present an interim OS analysis in 1L PD-L1 all-comer sq-NSCLC at the ASCO plenary on 31 May, 2026 – a slot usually reserved for data that are both statistically robust and According to the protocol, interim OS analyses are conducted under a Lan-DeMets alpha-spending framework with O’Brien-Fleming (OBF) efficacy boundaries under a group sequential analysis design, allowing interim OS analyses while preserving theoverall one-sided alpha of 0.025. As such, it prevents false positives from interim analyses while still allowing early stopping if As the exact planned interim OS information fraction (defined as planned interim OS event / planned final OS event) has notbeen formally disclosed,we use the protocal-specified 70% interim information fraction for PFS as a proxy for OSmaturityand construct a scenario table using multiple assumed planned final OS event counts. Under the classical OBF , where t is the interim OS information fraction the interim efficacy boundary is estimated at c. Z~2.34.We then apply AI-assisted Lan-DeMets spending adjustments toderiveZ~2.44 (Exhibit 1). Applying standard survival-analysis relationship: , where D is the planned interim OS events we estimate interim OS significance thresholds across varying OS maturity scenarios. Under these assumptions,interimOS significance would likely require an OS HR in the low-to-mid 0.7x range (Exhibit 2). Considering the trial timeline(estimated median follow-up at 21m with data cutoff Jan 2026 as reference; data cutoff may be between Jan to Mar), top two Long-tail “IO-like” benefit in HARMONi-A / HARMONi supports the biology, if not a home run In 2L+ EGFR-mutated NSCLC, both the China HARMONi-A trial and the global HARMONi trial showed only modest separationbetween AK112+chemo and chemo alone on OS – but importantly the small gap persists to the tail of the curves, consistentwith a long-tail survival benefit for the minority of patients who derive true immunotherapy benefit. In HARMONi-A, the final OS analysis (median follow-up ~32.5m) delivered an OS HR of 0.74 with AK112+chemo vs 14.1m forcontrol, and the treatment curve stays slightly above