Neoadjuvant Pembrolizumab Produces High pCR Rate With No Relapses at 3 Years in High-Risk dMMR/MSIHigh CRC

Nine weeks of single-agent neoadjuvant pembrolizumab (Keytruda) resulted in significant downstaging in patients with high-risk stage II/III mismatch repair-deficient/microsatellite instability(dMMR/MSI)–high colorectal cancer (CRC), no new adverse safety signals, and no disease relapses with nearly 3 years of follow-up, according to data from the phase 2 NEOPRISM-CRC trial (NCT05197322) presented at the 2026 AACR Annual Meeting.¹

At a median follow-up of 32.9 months (range, 26.4-41.4), the pathological complete response (pCR) rate among patients who underwent resection (n = 32) was 59%. Treatment was well tolerated, with grade 3 to 4 immune-related adverse effects (irAEs) occurring in only 2 patients (6.2%). Furthermore, at the March 31, 2026, data cutoff, no patients in the cohort (n = 31) experienced disease relapse.

Alongside the primary clinical finding, data from translational analyses indicated that ultrasensitive circulating tumor DNA (ctDNA) kinetics and baseline tumor T-cell receptor (TCR) clonality may be complementary predictors of pathological response, and postoperative ctDNA detection was concordant with RFS outcomes.

“These [translational data] are potentially practice-changing,” lead study author Yanrong Jiang, MA, MBBS, MRCP, stated in a presentation of the data. “Early ctDNA clearance may allow for the selection of super responders to single-agent immunotherapy for nonoperative management, postoperative ctDNA clearance allows for selective requirement of adjuvant therapies, and baseline TCR profiling has the potential to predict patient response to immunotherapy. We propose early adoption of these approaches into future prospective, adaptive, biomarker-informed clinical trials.”

Jiang is a physician in hematology/oncology training in the United Kingdom and a recent clinical research training fellow at the University College London Cancer Institute.

How was NEOPRISM-CRC designed?

NEOPRISM-CRC is a single-arm, translationally enriched clinical trial enrolling patients with radiologically defined high-risk stage II or III CRC. Eligible patients must have dMMR by immunohistochemistry or MSI-high by polymerase chain reaction and an ECOG performance status of 0 or 1.^1,2

At screening, all patients underwent trial CT CAP with or without MRI; colonic biopsies of tumor and normal tissue with FoundationOne CDx profiling; blood collection for germline testing, ctDNA, and peripheral blood mononuclear cells; and stool and oral sample collection. All patients received pembrolizumab at 200 mg intravenously every 3 weeks for cycle 1. After cycle 1, patients were stratified by tumor mutational burden (TMB): Those with TMB-high or TMB-medium disease received 3 additional cycles of pembrolizumab (cycles 2 and 3); patients with TMB-low disease proceeded directly to surgery. Surgery was performed 4 to 6 weeks after the last pembrolizumab cycle. Following surgery, patients could receive adjuvant FOLFOX or CAPOX at physician discretion.

Longitudinal plasma ctDNA was collected, and tumor tissue was assessed from a baseline biopsy and from the surgical resection specimen. ctDNA was analyzed using the NeXT Personal Dx tumor-informed molecular residual disease assay. The median limit of detection across all samples in the study was 1.54 parts per million (PPM; range, 1.09-3.58).

The primary end point was pCR rate. Secondary end points included 3-year relapse-free survival, overall survival, safety, and health-related quality of life. ctDNA response to neoadjuvant therapy, minimal residual disease monitoring, and genomic and microbiome biomarker signatures served as exploratory end points.

Of the 32 patients enrolled in NEOPRISM-CRC, 62% had locally advanced T4a or T4b tumors, and 34% had Lynch syndrome.¹ A baseline biopsy with whole genome sequencing and TCR profiling was performed in 31 patients. In the ctDNA biomarker analysis, 97 presurgical plasma samples from 25 patients were evaluable.

What did the ultrasensitive ctDNA analysis reveal?

The NeXT Personal Dx assay demonstrated 100% baseline ctDNA detection sensitivity across all patients assessed in the presurgical ctDNA cohort. The ctDNA detection range at baseline was 1.46 to 59,343 PPM, with a median of 678 PPM. Of all 67 ctDNA detections across the longitudinal presurgical time points, 24% fell within the ultrasensitive range of less than 100 PPM, highlighting the importance of the assay's depth of detection for capturing low-level residual disease that conventional platforms would miss.

Longitudinal ctDNA clearance progressed substantially across treatment cycles. In 25 patients with evaluable presurgical plasma data, the proportion with detectable ctDNA at a minimum threshold of 100 PPM declined from 92% at pre–cycle 1 (n = 25) to 60% at pre–cycle 2 (n = 25), 30% at pre–cycle 3 (n = 23), and 25% immediately before surgery (n = 24). The proportion with undetectable ctDNA rose correspondingly from 0% at pre–cycle 1 to 24% at pre–cycle 2, 43% at pre-cycle 3, and 58% before surgery. Post surgery (n = 25), 100% of patients had undetectable ctDNA.

Specifically, 24% of patients cleared ctDNA prior to cycle 2, and 58% achieved full ctDNA clearance by the end of neoadjuvant treatment. Notably, ultrasensitive monitoring detected persistent low-level ctDNA below 100 PPM in a subset of patients who would have been classified as ctDNA negative by conventional assays. Postsurgically, ctDNA-negative status at 4 to 5 weeks resulted in a 100% negative predictive value (NPV) for relapse.

How did ctDNA kinetics predict deep pathological responses?

An analysis of ctDNA kinetic profiles across the neoadjuvant treatment course revealed 3 distinct molecular responder phenotypes that mapped onto pCR outcomes:

• Super molecular responders (n = 6): ctDNA cleared before cycle 2 of pembrolizumab and remained undetectable through presurgery. All patients achieved a pCR.
• Dynamic molecular responders (n = 11): ctDNA levels fell rapidly across treatment cycles, reaching below 100 PPM presurgery. Nine patients achieved a pCR.
• Poor molecular responders (n = 8): ctDNA levels remained stable or rose above 100 PPM presurgery. No patients achieved a pCR.

The postoperative ctDNA surveillance data demonstrated perfect concordance with clinical outcomes across all 25 patients, with 100% specificity and 100% NPV. Two patients in the surveillance cohort received 3 months of adjuvant CAPOX.

What did the tumor TCR analysis show?

Baseline tumor TCR repertoire profiling was performed in 31 patients using RNA-based targeted sequencing (FUME-TCRseq). Baseline tumor TCR clonal fraction was higher in patients who achieved a pCR (n = 19) vs those who did not (n = 12; P = 3.735 × 10^–3). The pCR group also showed a substantially higher median large clonal fraction (defined as 0.001 < CF < 0.01) and a broader range of values extending to 0.7 compared with the non-pCR group.

What did the combined ctDNA and TCR Bayesian model demonstrate?

A Bayesian analysis integrating both the TCR clonal expansion data and the mean log change in ctDNA across treatment cycles (pre–cycle 2, pre–cycle 3, and presurgery vs pre–cycle 1) generated a sequential posterior probability landscape for pCR. Three key findings emerged from this analysis.

First, the Bayesian model confirmed that higher TCR clonal expansion was associated with greater average ctDNA reduction across treatment cycles, both of which were associated with a higher posterior probability of pCR. Second, dynamic ctDNA change added strong discrimination beyond TCR clonality alone, with the combination outperforming either biomarker in isolation. Third, the existence of distinct extreme-confidence regions in the probability landscape, encompassing patients with very high vs low predicted probability of pCR at either end, supports clinical consideration of divergent management strategies.

Disclosures: Jiang reported an advisory role for Flatiron Health.

References

1. Jiang Y, Militello A-M, Abbott C, et al. 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. Presented at: 2026 AACR Annual Meeting; April 17-22, 2026; San Diego, CA. Abstract CT170.
2. NEOadjuvant pembrolizumab in stratified medicine - colorectal cancer (NEOPRISM-CRC). ClinicalTrials.gov. Updated April 17, 2026. Accessed April 24, 2026. https://clinicaltrials.gov/study/NCT05197322