Certepetide, an NRP1-engaging tumor-penetration peptide, was halted in the BOLSTER cholangiocarcinoma trial

What was tried

Lisata Therapeutics ran NCT05712356, the Phase 2a BOLSTER study of certepetide, also known as LSTA1 and CEND-1, added to standard of care in advanced solid tumors with a focus on biliary tract cancer. The design was randomized, placebo-controlled, and quadruple-masked, with two cholangiocarcinoma cohorts. First-line patients received cisplatin, gemcitabine, and durvalumab, and second-line patients received a FOLFOX regimen, each with certepetide 3.2 mg/kg by slow intravenous push or matching placebo. Actual enrollment was 66 participants and the actual primary completion date was April 5, 2026 (ClinicalTrials.gov NCT05712356). The listed primary outcome was the number of subjects experiencing any adverse event, a safety measure. The study was terminated, with the posted reason that it was halted prematurely to allow analysis of the results (ClinicalTrials.gov NCT05712356).

The biological hypothesis

Certepetide is a cyclic tumor-penetration peptide of the iRGD class. It first binds alpha-v integrins on tumor vasculature, undergoes proteolytic cleavage that exposes a C-terminal R/KXXR/K motif, and that motif then engages neuropilin-1, NRP1, to trigger a transtissue transport pathway. The result is increased delivery of co-administered drugs deeper into tumor tissue without the peptide being chemically linked to those drugs. The mechanism is well described across reviews of iRGD biology (Thirumalai 2023). Related reviews describe the same biology (Yin 2017) and (Kang 2020) and the pharmacokinetics of the peptide have been characterized (Jarvelainen 2023). The clinical hypothesis was carried largely by pancreatic cancer data. A single-arm Phase 1b/2 of CEND-1 with nab-paclitaxel and gemcitabine reported encouraging response and survival figures that drew attention as a possible advance for a hard tumor (Dean 2022, doi:10.1016/S2468-1253(22)00167-4; Springfeld 2022, doi:10.1016/S2468-1253(22)00197-2).

What actually happened

BOLSTER moved the concept into a randomized, placebo-controlled setting in biliary tract cancer and stopped early to analyze results (ClinicalTrials.gov NCT05712356). Two features of the design stand out. The registered primary outcome was safety, the count of patients with any adverse event, not a controlled efficacy endpoint, which limits how strong an efficacy claim the trial can support on its own. The study also spread 66 patients across two cohorts and two arms, leaving small per-arm groups. Open Targets scores the NRP1 to cholangiocarcinoma association at 0.0898 overall, and that score is built almost entirely from literature evidence (0.739) with no genetic association support (0.000) (Open Targets Platform). NRP1 here is a delivery conduit rather than a genetically validated cancer driver, which is a different kind of target than the directly oncogenic nodes that anchor most successful oncology programs.

Failure mechanism, best guess

The most likely reading is translational mismatch. The preclinical and single-arm pancreatic data measured the effect of a penetration enhancer in settings that favored a visible signal, while the randomized biliary design asked the harder question of whether that enhancement produces a controlled benefit on top of active chemotherapy and immunotherapy. A penetration enhancer can raise intratumoral drug exposure and still fail to move endpoints if the backbone regimen is already near its efficacy ceiling or if biliary tumors do not gate on the NRP1 transport pathway the way the model systems suggested. The decision to stop and analyze, rather than to read out as planned, is consistent with an interim picture that did not clear the bar the single-arm data had implied. This is best classified as a translational mismatch between an enhancer mechanism with strong preclinical support and a controlled clinical setting that did not reproduce it.

How to prevent this next time

Two quantitative tools would have reduced the gap between expectation and result. First, an explicit power calculation tied to a controlled efficacy endpoint rather than a safety primary. With roughly 16 patients per arm in each cohort, a two sample comparison at alpha 0.05 has about 29 percent power to detect a standardized difference of 0.5 and about 40 percent at 0.6 (computed by normal approximation). A trial that small can describe safety but cannot reliably confirm or exclude an efficacy effect, so defining the minimum clinically relevant difference and powering to it would have made the readout interpretable. Second, a Bayesian posterior probability of success anchored to oncology base rates. The prior should be cautious, because single-arm signals translate weakly and oncology carries the lowest historical likelihood of approval among major areas, near 5.3 percent from Phase 1 (Wong 2019). Formally,

PP(success)=∫θ​P(final success∣θ,interim data)⋅π(θ∣interim data)dθ

Updating that prior with a biomarker enrichment strategy, selecting tumors with measurable NRP1 mediated transport, would have concentrated any true effect into a population where it could be detected. The single highest leverage change would have been ...

What this means for similar programs

Delivery-enhancement assets face a structural measurement problem. Their value is indirect, expressed through the drugs they accompany, so a randomized design has to separate the enhancer effect from the backbone effect with enough patients to do so. Programs built on iRGD and related peptides should pair a controlled efficacy endpoint with a transport biomarker from the start, and should be cautious about reading single-arm survival signals as evidence of a controlled benefit. The contrast between the pancreatic single-arm enthusiasm (Dean 2022, doi:10.1016/S2468-1253(22)00167-4) and a randomized biliary stop is the central lesson for the class.

Open questions

Did certepetide raise intratumoral exposure of the chemotherapy backbone in BOLSTER patients, and was that pharmacodynamic effect measured? Would a controlled efficacy endpoint, powered properly, have shown separation in first-line versus second-line cohorts? Is biliary tract cancer simply a poorer setting for NRP1 mediated transport than pancreatic cancer? The full analysis Lisata is running may answer the first two.

Sources

1. ClinicalTrials.gov, NCT05712356, certepetide BOLSTER Phase 2a in cholangiocarcinoma, enrollment 66, terminated, primary completion 2026-04-05, primary outcome adverse event count. Open Targets Platform, NRP1 to cholangiocarcinoma association 0.0898 (literature 0.739, genetic association 0.000). Dean A, et al. Lancet Gastroenterol Hepatol 2022,Springfeld C, et al. Lancet Gastroenterol Hepatol 2022,Thirumalai A, et al. Bull Cancer 2023,Yin H, et al. Mol Med Rep 2017,Kang S, et al. Polymers 2020,Jarvelainen HA, et al. Int J Mol Sci 2023,Wong CH, Siah KW, Lo AW. Biostatistics 2019,ChEMBL certepetide CHEMBL5315026, molecule type protein, max phase 2. FAERS queried for certepetide, no records (investigational compound). Power values computed by normal approximation. Available from: https://clinicaltrials.gov/study/NCT05712356.