Development and evaluation of a multi-cancer screening (MCS) test for cancers common in Asia

3052 Background: The majority (78%) of cancer deaths in Asia arise from cancers with no existing screening recommendations. Earlier cancer detection is associated with improved outcomes. While single-cancer screening by liquid biopsy (notably plasma EBV DNA for nasopharyngeal cancer) is recognized in Asia, multi-cancer screening (MCS) liquid biopsy testing is a potential opportunity to address this key need. We developed a ctDNA mutation-based testing workflow based on targeted amplicon-based next generation sequencing testing of plasma and matched buffy coat as required to exclude clonal hematopoiesis, with a custom pipeline incorporating clinico-demographic factors for tissue-of-origin (TOO) calling. Methods: Primary objectives were to evaluate assay performance for detection of cancer-associated alterations, as well as TOO prediction accuracy. Validation was performed on a sample cohort of 476 untreated patients, each with 1 of 10 different cancers (lung, liver, breast, colorectal, prostate, nasopharyngeal, pancreas, bile duct, acute and chronic myeloid leukemia). 50 initial age-matched healthy controls were also used. A weighted sum model was used for TOO prediction, which was evaluated in an expanded 1478-sample cohort. Weights assigned to each cancer-associated alteration and its co-occurring combinations represent the importance or contribution of each alteration to each potential TOO. The impact of inclusion of clinico-demographic data (where available) on TOO prediction was also evaluated. Results: Overall sensitivity for detection of cancer-associated alterations was 72.7% (346/476). Sensitivity is higher in metastatic cases (82.5%) compared to localized cases (41.8%). Sample-level specificity for cancer-associated alterations/EBV detection was 96% (48/50), also comparable to previous series. Cancer-associated alterations/EBV were detected in approximately 50% of localized cases for 3 cancer types with no established screening recommendations in Asia (lung - 54.6%; bile duct - 50.0%; pancreas - 48.6%). In the expanded cohort, a high confidence call for TOO was obtained in 55.4% (818/1478) of samples with positive findings. Clinico-demographic data inclusion as a model variable improved overall TOO prediction accuracy in 10 cancer types to 93.8% (767/818), compared to an accuracy of 90.2% without factoring in clinico-demographic data. Conclusions: The MCS ctDNA assay shows reasonable sensitivity and specificity for detection of cancer-associated alterations, based on a retrospective study cohort. Tumor TOO can also be correctly classified in the majority of patients. Incorporation of clinico-demographic data to personalize TOO recommendations also improved accuracy of calls. Further prospective studies and the addition of multi-omic capabilities will continue to be helpful to determine MCS performance in an Asian multi-ethnic environment.