Abstract
Background: Multi-cancer detection (MCD) tests can improve the efficacy of existing screening tests by detecting a greater number of cancers. For exerting a clinical impact, however, it is important that the MCD is able to accurately detect early-stage cancers, a feature that is limited in currently available tests. We had previously developed an alternate approach that combined serum metabolomics with machine learning-powered data analytics. Using this approach we have developed an MCD test that could concurrently detect 30 cancers with high accuracy. The goal of the present study was to clinically validate the performance of this test.
Patients and methods: A prospective, multicenter, observational study was conducted in which de-identified blood samples were collected from 10,074 participants with (n = 7246, 72%) and without (n = 2828, 28%) cancer. A blinded arm was also included to validate test performance. Sensitivity and specificity of cancer detection, and the accuracy of tissue of origin (TOO) identification was measured.
Results: The overall sensitivity obtained for cancer detection was 98.55% while the specificity was >99%. The assay demonstrated encouraging sensitivity for detecting early-stage cancers, although these findings require further validation in prospective screening studies, which ranged from 95% to 100% for different cancers. The overall sensitivity obtained for TOO was 95.62%, with accuracies ranging from 90% to 100% for early-stage cancers.
Conclusion: This study validates that our serum-based MCD test is uniquely capable of detecting early-stages of diverse cancers with high sensitivity and specificity, while also ascribing TOO with high fidelity. This test may potentially complement existing single-cancer screening approaches; however, prospective population-based implementation studies are required to establish its clinical utility, effectiveness, and impact on patient outcomes.
Clinical trial ID: CT/MD/2024/000007
Keywords
Cancer, Multi-cancer detection, Mass spectrometry, Metabolomics, Liquid biopsy, Machine learning