Atezolizumab continuation with lenvatinib or sorafenib versus lenvatinib or sorafenib alone in patients with hepatocellular carcinoma who progressed on first-line atezolizumab + bevacizumab. A test of whether continuing PD-L1 blockade after PD-L1 failure produces a different result than the first time.
IMbrave251 enrolls patients with hepatocellular carcinoma who progressed on first-line atezolizumab + bevacizumab and randomises them to receive atezolizumab + (lenvatinib or sorafenib) or (lenvatinib or sorafenib) alone. The experimental arm continues the same PD-L1 inhibitor that the patient's tumour has just escaped from, swapping the bevacizumab partner for a tyrosine kinase inhibitor. Whatever the tumour did to defeat atezolizumab in the first line is still in place when the second line begins.
The structural argument is that continuing checkpoint blockade after checkpoint failure does not address a new step in the immune cycle. The encounter step (PD-L1 release) was already engaged by the first-line therapy. The tumour has built its escape mechanism specifically against that step. Re-engaging the same step in second line is not adding coverage; it is repeating a step that has already failed. The TKI partner addresses a different mechanism (anti-angiogenesis) but does not change the structural status of the checkpoint axis.
This is not speculative. CONTACT-03 was the cleanest possible prior test of ICI rechallenge after ICI failure — atezolizumab + cabozantinib versus cabozantinib in renal cell carcinoma after prior checkpoint exposure. CONTACT-03 found identical PFS in both arms. The mechanism class (ICI rechallenge or continuation after ICI failure) has a 0% success rate in randomised Phase 3 settings to date. The framework's structural argument places IMbrave251 in the same class.
Each row below is a falsifiable claim with a value range. If the trial reports out within these ranges, the prediction is a hit. If it reports outside them in the named direction, the prediction is a miss.
| Metric | Predicted value |
|---|---|
| OS hazard ratio | 0.90–1.10 (not significant) |
| PFS hazard ratio | 0.90–1.10 (not significant) |
| ORR Δ (experimental − control) | 0–5 percentage points |
| Subgroup analyses | No subgroup with HR < 0.80 |
The critical test: if the experimental arm shows OS hazard ratio below 0.80 with statistical significance, the framework's claim that continuing checkpoint blockade after checkpoint failure cannot generate a new result is wrong, and this prediction is a miss. The class precedent (CONTACT-03) would also need explanation.
The structural reasoning behind the call — which steps in the immune cycle each drug covers, where the prediction would break, what it means for the design of follow-on trials in adjacent indications — is in the full report. The full report also includes the framework's per-metric reasoning for each row in the table above and the structural mapping for the trial's arms.
Fill the form to start the purchase. We reply within two business days with the price, scope, and a Stripe invoice. The full report is delivered as a PDF on payment.
* Trial cost estimate: ~554 patients × $40K median per patient (oncology Phase 3, US sites, Moore et al. 2018 cross-sectional analysis of 2015–2017 trials). Order-of-magnitude only — actual trial cost is not publicly disclosed.