REDMOD Finds Pancreatic Cancer Signs 475 Days Early — SkimNews
Get the Health newsletter
Daily health & science — research, biotech, public health, the studies worth knowing. Free.
- REDMOD identified subtle CT imaging signatures associated with pancreatic ductal adenocarcinoma an average of 475 days before clinical diagnosis in 219 patients whose scans had initially shown no evidence of cancer.
- REDMOD achieved 73% sensitivity, compared with 39% for experienced radiologists; on scans taken more than two years before diagnosis, its detection rate was 68% versus 23%.
- Gut published the multi-hospital comparison of 219 patients later diagnosed with pancreatic cancer and 1,243 controls who remained cancer-free for up to three years; 56 case scans, or 25%, preceded diagnosis by more than 24 months.
- REDMOD automatically outlined the pancreas and analyzed subtle texture patterns, eliminating manual organ tracing and reducing inconsistencies between specialists.
- REDMOD classified just over 81% of scans from an independent 539-patient group as cancer-free and 87.5% of the NIH-PCT dataset's 80 scans as cancer-free.
- REDMOD produced matching results on scans taken from the same patients several months apart in 90% to 92% of cases, which the researchers said strengthened evidence that it was identifying meaningful changes.
- REDMOD still needs prospective testing in high-risk groups, including people with unexplained weight loss or newly diagnosed diabetes; the researchers also flagged insufficient ethnic diversity in the study population.
Why it matters: REDMOD’s 475-day lead and 73% sensitivity, versus 39% for radiologists, offer pancreatic-cancer patients a treatment window before late-stage disease. The researchers’ key second-order point is that modeling cited in the study links an increase in localized cases from 10% to 50% with survival rates that would more than double.
Ask SkimNews
