评估用于识别新原发恶性肿瘤的分子肿瘤类型预测——一项真实世界临床基因组学分析

Background: Accurately distinguishing secondary malignancies from metastatic progression is critical in cancer surveillance, as each carries distinct prognostic and therapeutic implications. However, this distinction is often limited by overlapping radiologic features and the need for tissue biopsy. Methylation-based molecular tumor typing (MTT) of circulating tumor DNA (ctDNA) enables inference of tumor origin from blood and could support non-invasive and early identification of secondary malignancies. Method: The MTT algorithm was developed to differentiate methylation signals across 14 cancer types. For each evaluable sample, MTT reports top predictions and associated confidence scores. We evaluated concordance between MTT predictions and clinical evidence of secondary malignancies in samples tested with the Guardant Reveal (Guardant Health, Palo Alto, CA) assay for MRD detection. Insurance claims data from GuardantINFORM identified patients with evidence of multiple primary malignancies. Eligible patients had distinct primary cancers supported by sufficient ICD-10 codes and at least one evaluable Reveal sample within ±180 days of the second primary diagnosis. When multiple samples were available, the one closest to the second diagnosis was analyzed. To evaluate the effect of disease timeline on MTT performance, samples were stratified by interval since initial diagnosis: early (<=3 years) and late (>5 years). Second primary diagnoses in the early group were considered more likely to represent diagnostic refinement or recurrence, whereas those in the late group were presumed to reflect biologically independent new primaries. Results: Of the 101 eligible patients, 82 had a high confidence MTT prediction (81%). Among these 82, 88% (72/82) of top MTT predictions matched either the initial diagnosis or the second diagnosis. MTT results tended to reflect the new primary with longer intervals between initial and secondary diagnoses: For samples collected >5 years after the original diagnosis (late category), MTT matched the second primary in 68% (19/28) and the first in 14.3% (4/28) of cases. In comparison, among samples in the early category, MTT matched the second and first primary in 45% (19/42) and 47.6% (20/42) of cases, respectively. Among the 9 late cases whose MTT prediction did not match the second diagnosis, 4 had MTT aligned with the original primary, and the MTT predictions of another 3 were biologically related to the second diagnosis (e.g., uterine predicted as ovarian; colorectal predicted as gastroesophageal). Conclusions: This analysis demonstrates the potential of noninvasive epigenomic-based molecular tumor typing to identify evidence of new primary cancers during MRD assessment. This blood-based approach for MTT prediction could enable seamless integration of recurrence monitoring with the detection of new primaries.