Pattern Computer and MCW Report Early Results for a Three-Second Saliva Cancer Screen, With 91% Specificity and 44% Sensitivity

What "tier-1" screening means in this abstract

In the authors' framing, a tier-1 cancer screen is a point-of-care test that is "rapid, inexpensive, and highly specific to minimize unnecessary downstream workup." The classifier's operating threshold was therefore tuned toward specificity, and the release describes the result as discrimination maintained "under strict settings designed to eliminate false positives." Sensitivity was the price of that choice. At the reported threshold the scan flagged 44 of every 100 cancer-positive samples and cleared 91 of every 100 samples without cancer.

The comparison the authors draw is with centralized cell-free DNA methylation blood assays. Those, the abstract says, carry a roughly two-week turnaround and reagent costs, whereas the saliva scan returns a signal in seconds with no reagents. The abstract also states that "at comparable cohort scale, discrimination exceeded that reported for sequencing-based cfDNA methylation blood assays," without naming which assays or which published cohorts served as the comparator, so that claim cannot be checked from the text available.

Razelle Kurzrock's clinic supplied the patients and the labels

The study is led by Razelle Kurzrock, MD, FACP, director of the Center for Precision Oncology and Rare Cancers at the MCW Cancer Center and associate director of precision oncology at the Linda T. and John A. Mellowes Center for Genomic Sciences and Precision Medicine. Hui Zi Chen, MD, PhD, leads the arm in the lung cancer clinic, and Ann Maguire, MD, MPH, leads the arm in the Hereditary Cancer Risk Clinic. Cancer status and stage were abstracted from the electronic health record rather than adjudicated for the study, and the "no evidence of disease" groups were split by whether the patient was still on treatment. Kurzrock co-chairs the APE|APO Event Series meeting in Madison on August 27 and 28, 2026, and APO|APE News is owned by the organizer of that series.

Pattern Computer's contribution is the classifier, a Pattern Discovery Engine that outputs a symbolic equation in the spectral domain, which the company presents as an explainability feature, and which is also what let the team set an operating point for specificity. The abstract does not say how training and validation were split, whether the validation set was held out from the same 251 samples, or how the 16-patient on-treatment group was handled in the balanced-accuracy calculation.

A saturation curve pointing at 98%

The number the companies lean on comes from a statistical saturation analysis, which they say indicates that specificity above 98% is "likely attainable with a cohort of fewer than 500 samples." Mark R. Anderson, Pattern's chair and CEO, framed the next phase as chemistry rather than statistics: "Ongoing development is focused on attributing specific discriminative spectral signatures to underlying biological markers. The joint team can use orthogonal assays, including fractionation and targeted mass spectrometry, to identify the contributing analytes and host-response biomarkers."

Anderson also said the technology "shifts the diagnostic timeline from weeks to a sub-minute at the point of care, equaling traditional sequencing-based blood assays at scale." What the analytes are, and whether the spectral signal comes from tumor-derived material or from a host inflammatory response, is the open question the mass spectrometry work is meant to answer.

Why This Matters to the APO|APE Reader

A 44% sensitivity figure in a cohort where 99 of 251 participants already carried a cancer diagnosis, most of them drawn from a lung cancer clinic, is a case-finding result in a prevalent population, and the abstract reports nothing by stage. The blood-based multi-cancer tests this scan is measured against were judged on early-stage sensitivity across tens of thousands of asymptomatic people, and none of that has yet been attempted with saliva spectroscopy. The Anaheim poster and the JCO abstract are the entire public record so far, so a pathologist asked about this device by a hospital administrator can point to exactly one table of four numbers.