Next-generation multi-cancer early detection (MCED) assay: Independent blinded validation of a locked serum multi-omics NMR test with stage I sensitivity.

P Patrick C. Strasser (Whitehead Institute for Biomedical Research) M Maximilian Schneider (Wholomics GmbH, Ried Im Innkreis, Austria) N Norbert Thumfart (Wholomics GmbH, Ried Im Innkreis, Austria) H Hansjörg Habisch (Otto Loewi Research Center for Vascular Biology, Immunology and Inflammation, Medical University of Graz, Graz, Austria) T Tobias Madl L Luke Roy George Pike (Memorial Sloan Kettering Cancer Center, New York, NY)

Abstract

10536 Background: Blood-based MCED assays could complement existing screening, but many approaches show reduced sensitivity in early-stage disease. Wholomics developed a serum multi-omics MCED assay integrating metabolomic, lipidomic, and proteomic features quantified by proprietary NMR-based technology, designed to capture systemic biology associated with early tumor development. We report independent, blinded validation of a locked MCED assay developed on multi-center data; across training and validation, the program included >1,400 individuals and >13 cancer types. The full assay pipeline was locked before application to the independent validation cohort. Methods: Model development leveraged prior multi-center data (N = 1,211) from five European centers (669 cancers, 65% were stage 0-II; 442 controls plus precancerous/benign cohorts). We evaluated an independent validation cohort (N = 216) with 116 cancers (50% were stage 0-II [stage I n = 40; stage II n = 16]; cancers included colon, rectal, pancreatic, cervical, uterine, bladder, breast, gastric, renal, ovarian) and 100 controls with screening-relevant conditions (e.g., dyslipidemia, prediabetes, benign prostatic hyperplasia, osteoarthritis, etc.). Samples were analyzed with fully prespecified locked assay pipeline (proprietary NMR profiling according to SOP with predefined QC, data preprocessing, biomarker feature set, machine-learning model, and decision threshold) and applied to blinded outcome labels. Primary endpoints were specificity and sensitivity (overall, stage I, stage III-IV). 95% CIs for sensitivity and specificity were calculated using the Wilson score method; AUC 95% CIs were calculated using DeLong’s method. Results: The locked MCED assay achieved specificity of 99.0% (99/100; 95% CI 94.6–99.8; false-positive rate 1.0%) and overall sensitivity of 93.1% (108/116; 95% CI 87.0–96.5). Stage I sensitivity was 95.0% (38/40; 95% CI 83.5–98.6), Stage III-IV sensitivity remained high in advanced stages at 93.1% (54/58; 95% CI 83.6–97.3). AUC was 0.991 (95% CI 0.979-1). High overall sensitivity was observed across multiple tumor types, including pancreas, stomach, kidney, colon (each 100% overall) and rectum (92.3% overall). Conclusions: In an independent blinded validation, a locked serum multi-omics NMR-based MCED approach demonstrated high specificity and high sensitivity, including strong stage I performance, across multiple solid tumors in a screening-relevant case-control setting. These findings support clinical utility for shifting detection toward earlier, more treatable stages and warrant prospective studies to confirm real-world performance and downstream impact.

Article Details

Volume / Issue Vol. 44, Issue 16_suppl
Published June 01, 2026
Pages 10536-10536
ISSN 0732-183X
Publisher Lippincott Williams & Wilkins

Journal Info

Journal of Clinical Oncology

Lippincott Williams & Wilkins

ISSN: 0732-183X Health Sciences

Authors (6)

P

Patrick C. Strasser

Whitehead Institute for Biomedical Research

M

Maximilian Schneider

Wholomics GmbH, Ried Im Innkreis, Austria

N

Norbert Thumfart

Wholomics GmbH, Ried Im Innkreis, Austria

H

Hansjörg Habisch

Otto Loewi Research Center for Vascular Biology, Immunology and Inflammation, Medical University of Graz, Graz, Austria

T

Tobias Madl

L

Luke Roy George Pike

Memorial Sloan Kettering Cancer Center, New York, NY