The PROFOUND study: A large-scale prospective multicenter case-control trial of a multi-cancer early detection (MCED) test in China.

10541 Background: Early cancer detection reduces morbidity and mortality, but current recommended screening in China misses many lethal cancers. Emerging liquid biopsy-based MCED tests have the potential to increase the number of screen-detected cancers when added to recommended screening. In this study, we evaluate the performance of a MCED test tailored to 16 cancer types covering 85% of China's total cancer incidence, including malignancies without recommended screening tests in the large-scale, prospective, multicenter case-control PROFOUND study. Methods: A total of 10,541 participants from over 50 sites nationwide were enrolled, with the cohort comprised by a training set (7,487 participants: 4,448 cancer, 3,039 non-cancer) and a test set (3,054 participants: 1,740 cancer, 1,314 non-cancer). It included 16 major cancer types encompassing diverse pathological subtypes, with stage I-III cases accounting for 83.31% of all cancer samples. All cancer cases were confirmed via pathological diagnosis. Blood samples underwent targeted methylation profiling and were analyzed with the Genie-ADLA AI algorithm to train, validate and independently verify the MCED assay, as well as assess its clinical performance (sensitivity, specificity, tissue-of-origin [TOO] prediction accuracy). Results: In the validation set, specificity was 99.03% (95% CI: 98.27%–99.51%), overall sensitivity was 70.63% (95% CI: 67.23%–71.67%) and sensitivity for stage I–III cancers was 66.03% (95% CI: 63.53%–68.46%). The overall TOO prediction accuracy reached 85.21% (95% CI: 83.05%–87.20%). For the five highest-mortality cancers (liver, colorectal, esophageal, gastric and lung cancer), sensitivities were 87.85%, 90.48%, 83.78%, 72.85% and 69.54% respectively. Sensitivities of lung adenocarcinoma and non-adenocarcinoma were 48.75% and 92.96%, respectively. Among them, non-adenocarcinoma is more likely to be missed in recommended screening tests. For cancers without recommended screening (pancreatic, ovarian, gallbladder cancer, lymphoma), sensitivities reached 64.75%, 87.88%, 100% and 74.72%, respectively. Conclusions: Powered by the Genie-ADLA AI algorithm, the PROFOUND study's MCED assay demonstrated robust performance including high sensitivity, specificity and TOO prediction accuracy for 16 cancer types (including cancers do not have recommended screening). It may serve as a complementary tool to augment recommended screening and revolutionize early cancer detection paradigms. A prospective interventional clinical study covering 16 cancer types will be launched to further validate MCED's clinical effectiveness. Clinical trial information: NCT06217900 .