Scoring pre-registered predictions for the INTerpath-001 Phase 3 readout: Neoantigen immunogenicity reporting for intismeran autogene (mRNA-4157/V940) plus pembrolizumab in resected melanoma

Objective. We defined four levels of immunogenicity reporting across 12 clinical programmes. Among the 11 with patient-specific selection, six identified responding neoantigens and none linked those identities to clinical outcomes in the same patients. We predicted that the detailed INTerpath-001 Phase 3 melanoma readout would leave this relationship unresolved. A single qualifying analysis, at any cohort size, would have falsified our central prediction. Methods. We publicly posted six predictions and their falsification criteria on SSRN on 20 October 2026. The sponsors released the numerical results on 22 October, and the Madrid presentation followed on 24 October. The two scorers agreed on all six verdicts. Results. Under the authors' reading of P5, four predictions held, one failed and one was not evaluable. Under the alternative reading, three held, two failed and one was not evaluable. P3 had been withdrawn before deposit. The materials reviewed reported no qualifying Phase 3 immune response to an encoded neoantigen and no analysis linking responding targets to patient outcomes. The selector-disclosure verdict depends on whether one slide disclosed a new ranking criterion for the intismeran selector; both tallies use the same prespecified falsifier. Tumour-mutational-burden subgroup estimates were unavailable. The primary recurrence-free survival hazard ratio was 0.58 (95% CI 0.45–0.75), below our forecast of 0.60–0.78. Before the readout, we identified this forecast as most vulnerable and explained why an interim efficacy analysis could yield a lower estimate. The 247 observed events fell within the 200–500-event category discussed in advance. Interpretation. Adding intismeran to pembrolizumab improved recurrence-free survival. Our framework anticipated the reporting gap that would remain at this readout. Phase 2b investigators had specified an exploratory antigen-specific T-cell endpoint. Whether comparable per-target measurements were obtained in Phase 3, and whether interim timing contributed to the numerical miss, remain open. A further prediction, that the full trial report will contain no L4 analysis, was registered publicly on OSF Registries on 6 November 2026 (DOI 10.17605/OSF.IO/VK6RW; section 7).