Eciskafusp alfa (PD1-IL2v): a Phase 1 solid-tumor program stopped at a 4.4% predicted chance of clearing its ORR gate

What was tried

Eciskafusp alfa (RO7284755, also RG6279) is a bispecific immunocytokine that fuses an anti-PD-1 antibody to an interleukin-2 variant. The IL-2 variant is engineered to signal through the intermediate-affinity IL-2 receptor beta and gamma chains while avoiding the high-affinity alpha chain (CD25), so the molecule is described as a "betagamma-only" IL-2v (Codarri Deak et al., Nature 2022). The anti-PD-1 arm delivers that cytokine in cis to PD-1 positive T cells.

Roche ran NCT04303858 as an open-label, randomized, parallel Phase 1A/1B study in adults with locally advanced or metastatic solid tumors who had exhausted standard care. The study tested eciskafusp alfa alone and in combination with the anti-PD-L1 antibody atezolizumab, with dose-limiting toxicity, adverse events, recommended dose, and investigator-assessed objective response rate as primary measures (ClinicalTrials.gov NCT04303858). Enrollment reached 189 participants. The study started on 4 May 2020 and reached its primary completion on 2 October 2025. Roche terminated it and recorded the reason verbatim: a risk/benefit analysis put the probability of meeting the prespecified objective response rate gate at 4.4% (ClinicalTrials.gov NCT04303858).

The biological hypothesis

The premise was that PD-1 blockade alone leaves many tumors unresponsive because the stem-like CD8 T cell pool that sustains an anti-tumor response is not adequately expanded. An IL-2 signal delivered in cis to PD-1 positive cells was shown preclinically to push stem-like TCF-1 positive CD8 T cells toward better effectors, producing superior control in chronic LCMV infection and in tumor models (Nature 2022). Combining PD1-IL2v with anti-PD-L1 broke resistance in mouse models by enriching stem-like tumor-reactive CD8 T cells (Immunity 2023), and a 2025 human lung cancer translational study reported a multifaceted T cell response after PD-1-targeted cis delivery of the IL-2 variant (Science Translational Medicine 2025).

The same IL-2Rbetagamma-biased variant had already been carried into two earlier Roche immunocytokines, the CEA-targeted cergutuzumab amunaleukin and the FAP-targeted simlukafusp alfa (Oncoimmunology 2017). Open Targets assigns IL2RB an overall association of 0.51 with neoplasm, with the signal coming from literature (0.92) and clinical (0.80) evidence rather than human genetics, which records no genetic association datatype for this pair (Open Targets Platform, ENSG00000100385).

What actually happened

The trial enrolled to 189 and ran for more than five years before Roche stopped it. No results were posted to the registry. The decisive number was internal and prospective: at the planned final analysis, the modeled probability of clearing the objective response rate gate was 4.4% (ClinicalTrials.gov NCT04303858). That figure is a statement about expected efficacy, not toxicity or recruitment. The program was discontinued because the response signal was not tracking toward the bar the sponsor had set, even with atezolizumab added.

Failure mechanism, best guess

The most consistent reading is a translational efficacy shortfall in the same place the IL-2 receptor agonist class keeps stumbling. Preclinical models show large, durable expansions of stem-like CD8 T cells, yet the clinical response rate in unselected, checkpoint-experienced solid tumors stays low. The betagamma-biased design was meant to fix the CD25 driven liabilities of native IL-2, including regulatory T cell expansion and vascular toxicity, and it appears to have produced a tolerable molecule that still did not convert into responses at the rate required. A 2025 report on an IL-15 based cis-targeted competitor argued for superior antitumor activity over PD1-IL2v in preclinical comparison (Journal for ImmunoTherapy of Cancer 2025), which is consistent with the view that the IL-2v cis-delivery approach, as configured here, sat below the activity threshold needed in this setting.

How to prevent this next time

Two quantitative levers would have changed the decision calculus earlier. First, a formal Bayesian predictive probability of success run at each dose-finding readout, rather than only at the final analysis, would have surfaced the 4.4% trajectory while the program was smaller. The predictive probability is the integral of the final-success probability over the posterior on the treatment effect:

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

A prespecified rule to stop expansion when this quantity fell below a threshold such as 10% would have ended cohorts before they reached 189 enrolled. Second, historical base-rate adjustment should have anchored expectations: oncology assets entering Phase 1 reach approval roughly 5.3% of the time (BIO, Informa Pharma Intelligence, and QLS Clinical Development Success Rates 2011-2020), and the broader estimate of overall probability of success sits near 13.8% across indications (Wong, Siah, and Lo, Biostatistics 2019). A program whose own interim modeling lands at 4.4% is well below even that low base rate, which argues for biomarker enrichment toward PD-1 positive, TCF-1 positive stem-like T cell content before expansion rather than enrolling unselected solid tumors. The single highest leverage change would have been a prespecified Bayesian predictive-probability futility gate at the dose-finding stage tied to a PD-1 positive stem-like CD8 enrichment biomarker, so the 4.4% trajectory triggered a stop at tens of patients rather than 189.

What this means for similar programs

The IL-2 receptor agonist field has now produced a consistent pattern: compelling expansion biology, manageable safety after CD25 avoidance, and response rates that do not clear go criteria in unselected solid tumors. Programs built on IL2RB or IL2RG agonism should treat preclinical effector expansion as necessary but not sufficient, and should price in the documented gap between expansion and response. Cis-targeting through a tumor or T cell antigen sharpens delivery but does not by itself guarantee the response rate, as this Phase 1A/1B result shows.

Open questions

Did response correlate with on-treatment stem-like CD8 expansion in the human samples, and if so at what threshold? Would a CD25-engaging or IL-15 based payload have cleared the gate where the betagamma-only IL-2v did not? Was the atezolizumab combination subgroup meaningfully different from monotherapy at the doses tested? The unposted response data leave each of these unresolved.

Sources

1. • ClinicalTrials.gov, NCT04303858, study record, status TERMINATED, enrollment 189, termination reason quoting the 4.4% predicted probability of meeting the ORR gate. - Codarri Deak et al., "PD-1-cis IL-2R agonism yields better effectors from stem-like CD8 T cells," Nature 2022,- Tichet et al., "Bispecific PD1-IL2v and anti-PD-L1 break tumor immunity resistance," Immunity 2023,- "PD-1-targeted cis-delivery of an IL-2 variant in human lung cancer," Science Translational Medicine 2025,- Klein et al., "Cergutuzumab amunaleukin (CEA-IL2v)," Oncoimmunology 2017,- "PD-1-targeted IL-15 mutein SOT201 superior to PD1-IL2v," Journal for ImmunoTherapy of Cancer 2025,- Wong, Siah, Lo, "Estimation of clinical trial success rates," Biostatistics 2019,- BIO, Informa Pharma Intelligence, QLS Advisors, Clinical Development Success Rates 2011-2020 (oncology Phase 1 likelihood of approval near 5.3%). - Open Targets Platform, target IL2RB (ENSG00000100385), neoplasm association 0.51. - Claidex Mechanism Risk Score, IL2RB, computed live from the Claidex claims table. Available from: https://clinicaltrials.gov/study/NCT04303858.