Validation
Validation against published Phase 3 trial results.
Evinexus emulated the treatment and control arms of CheckMate-057 and TAILOR, then compared the digital twin estimates with the published trial estimates and confidence intervals. Informed by the FDA-collaborative ENCORE research approach, Evinexus assesses whether each estimate lands inside the trial's confidence interval (estimate agreement) and preserves the direction of the treatment effect (direction agreement). Across both case studies below, both criteria were met on every comparison — while removing years of follow-up.
Concordance
Published trial and digital twin estimates.
Each row is one arm's estimate. The reference trial's own 95% CI is shaded behind every row, so an estimate landing inside that band reads as agreement rather than leaving the reader to compare two intervals by eye.
Case 1 · Methodology study
CheckMate-057 reproduced with an Evinexus digital twin control, 194 days sooner to the primary analysis.
Overall survival · advanced non-squamous NSCLC
Reproduced hazard ratio
0.73
vs trial 0.73 · all three methods within CI
Study duration shortened
194d
up to 6.4 months sooner to primary OS analysis
Methods at full regulatory + estimate agreement
3 / 3
Three prespecified borrowing approaches
All three twin estimates reproduce the trial's hazard ratio of 0.73 and fall inside its 95% confidence interval.
Case 1 · Methodology study — reference trial vs. Evinexus digital twin
Reference RCT
Evinexus digital twin
RWD study
Hazard ratio (95% CI), nivolumab vs control
Case 1 · Methodology study. Reference trial and Evinexus digital twin estimates, hazard ratio with 95% confidence interval, by arm.
Reference RCT, CheckMate-057 (original RCT). Hazard ratio 0.73 (0.59–0.89).
Evinexus digital twin, Method 1. Hazard ratio 0.73 (0.59–0.90).
Evinexus digital twin, Method 2. Hazard ratio 0.73 (0.59–0.89).
Evinexus digital twin, Method 3. Hazard ratio 0.73 (0.60–0.91).
RWD study, Method 1. Hazard ratio 0.73 (0.59–0.92).
RWD study, Method 2. Hazard ratio 0.74 (0.61–0.91).
RWD study, Method 3. Hazard ratio 0.72 (0.59–0.88).
Case 2 · RAS WT metastatic colorectal cancer
Cetuximab + FOLFOX-4 vs FOLFOX-4 alone
TAILOR · Progression-free survival (PFS)
Methods reproduced
3 / 3
Three prespecified borrowing approaches
Days saved
967d
Method 1
Hybrid cohort
473 patients
193 randomized + 280 control, of which 80 digital twins
Every method's 95% CI overlaps the trial reference.
Case 2 · RAS WT metastatic colorectal cancer — reference trial vs. Evinexus digital twin
Reference RCT
Evinexus digital twin
Hazard ratio (95% CI), cetuximab vs control
Case 2 · RAS WT metastatic colorectal cancer. Reference trial and Evinexus digital twin estimates, hazard ratio with 95% confidence interval, by arm.
Reference RCT, TAILOR (original RCT). Hazard ratio 0.69 (0.54–0.89).
Evinexus digital twin, Method 1. Hazard ratio 0.74 (0.58–0.94).
Evinexus digital twin, Method 2. Hazard ratio 0.73 (0.56–0.92).
Evinexus digital twin, Method 3. Hazard ratio 0.73 (0.57–0.94).
Twin estimates are propensity-weighted. Full results at SABCS 2026 and ASH 2026; prospective and regulatory validation warranted.
Regulatory context
Regulatory and study-design context.
FDA
Considerations for the Design and Conduct of Externally Controlled Trials for Drug and Biological Products (draft guidance, 2023).
FDA ENCORE
An FDA-collaborative research project developing methods to calibrate real-world evidence against randomized trials. Not an FDA qualification or acceptance framework.
ICH E10
Choice of Control Group and Related Issues in Clinical Trials.
Where it degrades
Limitations
These results are retrospective emulations of completed trials. Prospective and regulatory validation is required before applying the approach to an active development program.
Want to review the full validation results?
Contact us