Abstract Background: Patients (pts) with operable, high risk stage II-III dMMR/MSI colorectal cancer (CRC) achieved a pathological complete response (pCR) of 59% in the phase 2 NEOPRISM-CRC trial of neoadjuvant pembrolizumab (pembro). Here we report longitudinal tumor informed ctDNA dynamics and T cell receptor (TCR) clonality as a marker of treatment response, minimal residual disease, and updated survival outcomes. Methods: 32 pts with newly diagnosed radiologically node positive or high risk T3/T4 dMMR/MSI CRC received 1 cycle of pembro 200 mg IV Q3W. 31 pts with TMB-medium (6-19 mutations/Mb) or TMB-high (TMB-H) (≥20 mutations/Mb) tumors, evaluated by FoundationOne®CDx, received 2 further cycles. Surgery was performed 4-6 weeks after the final cycle. Primary endpoint was pCR. Secondary endpoints included 3-year relapse free survival, overall survival and safety. ctDNA profiling was performed using the whole genome tumor-informed Personalis NeXT Personal® assay (LOD95 of 3. 45 PPM), tracking up to 1800 patient-specific variants over 11 timepoints; at baseline, prior to each cycle, pre and post-surgery, and at 3, 6, 9, 12, 24, 36 months (mo) follow up. RNA-based TCR sequencing (FUME-TCRseq) was conducted on pre-treatment tumor biopsies. Unique CDR3 amino acid sequences were quantified to derive TCR repertoire metrics. Repertoires were stratified by clonotype size and compared between histopathological responders and non-responders. Results: 23/31 (74%) pts with TMB-H tumors had evaluable paired tissue and plasma samples (n=207). Baseline ctDNA was detected in all pts (23/23, 100%). Pre-surgery ctDNA clearance was strongly associated with pCR in 13/14 pts (PPV=92. 9%) while persistent pre-surgical ctDNA detection predicted residual disease in 8/9 pts (NPV=88. 9%). 6/6 pts who cleared ctDNA prior to cycle 2 achieved pCR (PPV=100%). All pts had undetectable ctDNA up to 24 mo post-surgery. Tumour TCR analysis demonstrated that large clonotypes (relative abundance 0. 001-0. 01) occupied significantly greater clonal space in pts with pCR compared with non-pCR (p=0. 011). In a Bayesian model incorporating ctDNA kinetics and baseline TCR clonality, combined multimodal profiling was highly predictive of response with posterior probabilities ≥0. 7 identifying pts with pCR (14/14, PPV 100%). At data cut off of 1st November 2025, median follow up was 28. 7 mo (23. 0-36. 5) with no disease relapse. Conclusions: Ultrasensitive longitudinal tumor-informed ctDNA monitoring may dynamically predict response to neoadjuvant immunotherapy. Expanded intratumoral TCR clonality in responders highlights the immunobiological basis of pathological response which may aid early patient stratification and guide de-escalated neoadjuvant strategies. Citation Format: Yanrong Jiang, Anna-Maria Militello, Charles Abbott, Bailiang Li, Monika Madrova, Mark Saunders, Jenny Seligmann, Timothy Iveson, Richard Wilson, Janet Graham, Sandra Irvine, Rubina Begum, Reshma Bhat, Gabriel Naylor-Leyland, Andrew Plumb, Austin Obichere, Evelyn Y. Wong, Sean Boyle, Richard Chen, Manuel Rodriguez-Justo, Marnix Jansen, Kai-Keen Shiu. Neoadjuvant pembrolizumab stratified by tumor mutation burden in high-risk stage II-III dMMR/MSI colorectal cancer (NEOPRISM-CRC): Perioperative ultrasensitive ctDNA monitoring and tumor-infiltrating TCR repertoire for treatment response prediction abstract. In: Proceedings of the American Association for Cancer Research Annual Meeting 2026; Part 2 (Late-Breaking, Clinical Trial, and Invited Abstracts) ; 2026 Apr 17-22; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86 (8Suppl): Abstract nr CT170.
Building similarity graph...
Analyzing shared references across papers
Loading...
Jiang et al. (Fri,) studied this question.
www.synapsesocial.com/papers/69e473bd010ef96374d8f75c — DOI: https://doi.org/10.1158/1538-7445.am2026-ct170
Yanrong Jiang
Anna-Maria Militello
Charles W. Abbott
Cancer Research
University College London
University of Glasgow
Cancer Research UK
Building similarity graph...
Analyzing shared references across papers
Loading...