An Exosome-Based Liquid Biopsy for the Detection of Early-Onset Colorectal Cancer: The ENCODER Multicenter Study

Chandrika Kurpad MS, LCGC - CGA-IGC Communications Committee Member

Colorectal cancer used to be thought of as a disease of older adults. That assumption no longer holds. Rates of colorectal cancer in people under 50 — known as early-onset colorectal cancer (EOCRC) — have climbed steadily for decades. Today it is the leading cause of cancer-related death in young men and the second in young women. The trend is global, documented across North America, Europe, and Asia.

The problem compounds itself: young people are not getting screened. Despite guidelines recommending colonoscopy from age 45, compliance among younger adults sits below 20%. Many find the procedure invasive, inconvenient, or simply not on their radar. By the time symptoms appear, the cancer has often reached an advanced, harder-to-treat stage.

A new study published in Gastroenterology offers a potential answer: a blood test developed specifically for this younger population, with accuracy metrics that rival or surpass existing approaches for detecting early-stage disease.

The Core Idea: miRNAs in Blood as Cancer Signals

When cancer grows, it leaves molecular fingerprints in the bloodstream. Among the most informative are microRNAs (miRNAs) — tiny RNA molecules that regulate gene expression. Cancer cells shed miRNAs in two forms: floating freely in plasma (cell-free miRNAs), and packaged inside protective bubbles called exosomes.

Exosomes are particularly attractive as biomarker vehicles. Their lipid membrane shields their RNA cargo from degradation, giving exosomal miRNAs a stability advantage over bare cell-free molecules. The ENCODER study combined both types, hypothesizing that together they would outperform either alone.

This builds on prior work from the same group — a 4-biomarker cell-free test achieving ~88% AUROC. ENCODER is the next generation: more biomarkers, exosomal enrichment, and a more powerful machine learning model.

How the Test Was Built

The study followed the Early Detection Research Network (EDRN) framework — a rigorous multi-phase pipeline designed to prevent overfitting and ensure results replicate in truly independent data.

1.  1  Discovery — RNA Sequencing (n = 118)

Small RNA sequencing identified candidate miRNAs differentially expressed in EOCRC versus healthy controls. A multi-step filtering process — statistical significance, fold-change thresholds, tissue concordance, external validation, and LASSO regression — narrowed hundreds of candidates to 6 final biomarkers.

2.  2  Training — Machine Learning (n = 192)

The 6 miRNAs were measured by RT-qPCR in a clinical cohort. An XGBoost model was trained and hyperparameters tuned. After training, the model was locked — no further adjustments permitted.

3.  3  External Testing — Independent Prospective Cohort (n = 191)

The locked model was applied to a completely separate cohort from different institutions across four countries — the gold standard in biomarker validation.

4.  4  Post-Surgical Monitoring (n = 41)

Matched pre- and post-surgery samples assessed whether the test tracks tumor burden — a critical property for any biomarker intended for clinical use.

The Final Biomarker Panel

Six miRNAs made it through the entire selection pipeline. Two are cell-free; four are exosomal. The primary drivers of accuracy were exo-miR-32-5p and cf-miR-625-3p, with the remaining four contributing finer discrimination. All six have prior biological relevance in colorectal carcinogenesis.

Cell-free miRNAs

 hsa-miR-625-3p     hsa-miR-30d-5p 

Exosomal miRNAs

 hsa-miR-32-5p     hsa-miR-486-3p     hsa-miR-550a-3-5p     hsa-miR-625-3p

Notably, these markers appear specific to early-onset disease. When tested against four independent cohorts of patients over 50, their discriminatory ability was substantially lower — consistent with the hypothesis that EOCRC is biologically distinct from later-onset CRC.

The Results

In the independently tested cohort, ENCODER achieved performance that is striking for a blood-based cancer screening test

The most clinically meaningful figure: for stages I–III — the window where curative treatment is possible — sensitivity reached 97.3%. Stage II disease was detected with 100% sensitivity.

How does ENCODER compare to the current FDA-approved blood-based CRC screening test?

Test Levels Fall After Surgery

A critical property of any cancer biomarker is whether it actually tracks tumor biology. After curative-intent surgery in 41 patients, ENCODER risk scores dropped significantly — reaching negative in samples collected 4 or more days post-operatively.

In the first 1–3 days post-surgery, scores fell modestly; by days 4–7, the decrease was substantial enough that most patients crossed into the “negative” range. This kinetic behavior supports the interpretation that the biomarkers are tumor-derived and specific — not general markers of illness. It also opens the door to future applications in minimal residual disease monitoring, though the authors note that a single post-operative timepoint per patient is insufficient to draw conclusions about recurrence surveillance.

Limitations and Open Questions

• Sample sizes, while the largest EOCRC cohort to date, remain modest — particularly for stage-specific estimates, leading to wide confidence intervals in some subgroups.

• The study enrolled predominantly White and Asian participants. African Americans, who face disproportionately high EOCRC mortality, were not well represented and warrant dedicated future study.

• The exosome isolation method is precipitation-based and does not yield pure exosomes — a practical tradeoff that favors RNA yield and clinical scalability over absolute purity.

• Head-to-head trials against stool-based RNA tests and real-world implementation studies in healthcare systems are still needed.

• Cost-effectiveness analyses have not yet been performed.

The Bottom Line

ENCODER represents a meaningful advance in colorectal cancer screening for young adults. By combining cell-free and exosomal miRNAs with a powerful machine learning classifier, the assay achieves sensitivity and specificity that exceed current blood-based alternatives, remains stable across age groups, and requires only a small plasma volume — making it practically translatable to clinical settings.

It is not a replacement for colonoscopy, but it doesn’t need to be. A blood test that can be drawn during a routine checkup, flag high-risk individuals for confirmatory colonoscopy, and potentially catch pre-cancerous lesions — that is a complementary tool with real clinical value in a population that currently avoids screening altogether.

CITATION

Mannucci A, Balaguer F, Yamada Y, Nagasaka T, Toiyama Y, Okugawa Y, Martí-Gallostra M, Jiménez-Toscano M, Vidal-Tocino R, Jiménez F, Perea J, Quintero E, Boland CR, Cavestro GM, Goel A; on behalf of the SECOC-ENCODER Collaborators. An Exosome-Based Liquid Biopsy for the Detection of Early-Onset Colorectal Cancer: The ENCODER Multicenter Study. Gastroenterology. 2026;170(2):330–343. doi:10.1053/j.gastro.2025.08.013. ClinicalTrials.gov NCT06342401.