REDMOD Spots Pancreatic Cancer Signs 475 Days Early — SkimNews
Get the Health newsletter
Daily health & science — research, biotech, public health, the studies worth knowing. Free.
- REDMOD was tested on pre-diagnosis CT scans from 219 people later diagnosed with pancreatic ductal adenocarcinoma and 1,243 matched people who remained cancer-free; it identified cancer-associated signatures an average of 475 days before clinical diagnosis.
- REDMOD achieved 73% sensitivity on the 219-patient case set, versus 39% for experienced radiologists; on scans taken more than two years before diagnosis, its detection rate was 68% versus their 23%.
- REDMOD correctly classified just over 81% of scans as cancer-free in an independent 539-patient group and 87.5% in the 80-patient NIH-PCT dataset.
- REDMOD produced matching classifications on 90% to 92% of scans repeated months apart from the same patients, showing that its results were consistent over time.
- REDMOD automatically identified and outlined the pancreas, removing the need for specialists to manually trace its boundaries and reducing inconsistencies between tracers.
- REDMOD still needs prospective testing in high-risk groups—including people with unexplained weight loss or recently diagnosed diabetes—before routine clinical use; researchers also flagged insufficient ethnic diversity in the study cohort.
- REDMOD's investigators said the framework identified imaging signatures of stage 0 pancreatic ductal adenocarcinoma in pancreatic tissue that appeared normal, framing the result as pre-clinical interception.
Why it matters: For patients who receive late pancreatic-cancer diagnoses, survival is the stake: cited modeling says increasing localized cases from 10% to 50% would more than double survival rates. REDMOD’s 73% sensitivity versus radiologists’ 39% provides a concrete detection advantage, though prospective validation in high-risk groups remains outstanding.
Ask SkimNews
