Blood omics is now sensitive enough that the main problem is not whether blood contains disease-relevant information, but claim escalation. A circulating signal that classifies cases and controls may be written as diagnosis, mechanism, burden, prognosis, response prediction, or readiness before the sampled compartment has been asked what inference it can bear. This Review proposes claim-first adjudication for circulating omics in heterogeneous disease. The rule fixes intended claim, sampled compartment, source proximity, cohort mixture, coupling evidence, recurrence unit, falsifier, permitted wording, prohibited wording, and decision consequence before accuracy is interpreted. A reproducible evidence-atlas protocol and reference-audit workbook show why the sequence changes verdicts: Alzheimer blood biomarkers support comparator-defined pathology aid, ctDNA supports source-linked material detection, host-response signatures recur as modules or trajectories, and high-mixture signals require demotion or redirection rather than diagnostic inflation. Validation should begin by asking what blood can claim. This manuscript is a preprint and has not undergone peer review. Subsequent peer-reviewed versions may differ.
han et al. (2026) studied this question.