OncoFirm™ / Cancer Atlas 2026 / Multi-biomarker analysis
Cancer Atlas 2026 · Multi-biomarker analysis
Multi-Biomarker Analysis for Cancer Detection: Panels, Algorithms and Research Standards in 2026
One tumor marker is rarely accurate enough on its own. Combining markers in a single locked algorithm can help, but only when the gain survives independent testing. This Atlas chapter covers why specificity dominates at population scale, which multi-marker tests are FDA cleared or approved, how panels should be validated, and the limits of multiplex lateral flow.
At a glance
Why combine
One combined score can raise sensitivity; separate cut-offs for each marker add false positives
Cleared or approved
ROMA, OVA1, Overa (pelvic mass); PHI, 4Kscore (prostate biopsy decisions); Cologuard Plus, ColoSense (stool-based colorectal screening)
Not FDA approved
Multi-cancer blood tests, including protein plus DNA panels, as of 1 October 2026
Standards
Pepe phases, PRoBE, TRIPOD+AI, STARD-AI and REMARK; lock-down, external validation and calibration
OncoFirm™ status
MultiDx is a research concept; research use only; no screening or multi-cancer claims; not for sale
On this page
Key numbers
Multi-biomarker analysis in three numbers
19.4%positive predictive value of a protein plus DNA blood test in DETECT-A, despite 98.9% specificity
27main checklist items in TRIPOD+AI (2024) for reporting prediction model studies
3–4test lines per lateral flow strip before readability suffers (2025 review)
Sources: Lennon et al., Science 2020, as summarized by The ASCO Post; Collins et al., BMJ 2024; Velotta et al., Sensors 2025.
Rationale
Why combine markers at all?
Sensitivity is the share of people with cancer who test positive. Specificity is the share of people without cancer who test negative. Positive predictive value (PPV) is the share of positive results that are true cancers. Single tumor markers rarely reach the specificity needed when cancer is rare. See Cancer biomarkers explained in the OncoFirm™ Cancer Atlas 2026.
A panel combines several markers, often with age or other clinical data, in one algorithm that gives one score. This can raise sensitivity. In its pivotal study, the five-protein OVA1 test detected 92% of ovarian cancers, against 74% for CA-125 alone. The price was a specificity of 54% and a PPV of 31%.
Suppose seven independent markers each have 98% specificity, and any single positive counts. Combined specificity falls to about 87% (0.98 multiplied by itself seven times; our arithmetic). That is why a sound panel uses one combined score and one threshold.
Population scale
Why specificity dominates at population scale
PPV depends on prevalence, the share of tested people who have the disease. The US National Cancer Institute states the rule: for a given sensitivity and specificity, the lower the prevalence, the lower the PPV. In screening, most people do not have cancer, so a small false-positive rate still produces many false alarms.
Worked example (our arithmetic, not a study result). Screen 100,000 people. Assume 0.5% have cancer (500 people) and the test finds 60% of them (300 true positives). Only specificity changes in the table below.
| Specificity | False positives | True positives | PPV |
|---|---|---|---|
| 92% | 7,960 | 300 | 3.6% |
| 97.4% | 2,587 | 300 | 10.4% |
| 99.5% | about 498 | 300 | 37.6% |
In DETECT-A, about 10,000 women aged 65 to 75 had a protein plus DNA blood test. Blood testing alone had 98.9% specificity and a PPV of 19.4%. Adding a confirmatory PET-CT scan raised specificity to 99.6% and PPV to 28.3%. This is why the cleared blood algorithms below are meant for people already at raised risk, such as women with a pelvic mass.
Established
Cleared and approved multi-marker tests, and what each is for
As of 1 October 2026, a small group of multi-marker tests has FDA clearance or approval, each for a narrow use.
| Test | What it combines | FDA status | Intended use |
|---|---|---|---|
| ROMA | HE4, CA-125, menopausal status | Cleared 2011 | Pelvic (adnexal) mass before surgery; not a screening test |
| OVA1 | CA-125, four other proteins, menopausal status | Cleared 2009 | Adnexal mass; referral decisions |
| Overa | CA-125, HE4, FSH, two other proteins | Cleared 2016 | Adnexal mass; referral decisions |
| PHI | PSA forms, including p2PSA | Approved 2012 | Men 50 and older, PSA 4.0–10.0 ng/mL, non-suspicious rectal exam |
| 4Kscore | Total, free and intact PSA, hK2, plus age, exam and biopsy history | Approved 2021 | Risk of Gleason 7 or higher cancer; one laboratory site |
| Cologuard Plus | Stool DNA markers plus hemoglobin | Approved 2024 | Colorectal screening, average-risk adults 45 and older |
| ColoSense | Stool RNA markers plus hemoglobin | Approved 2024 | Colorectal screening, average-risk adults 45 and older |
The two stool tests are the main FDA-approved multi-analyte screening tests. Their makers report colorectal cancer sensitivity of 95% for Cologuard Plus (at 94% specificity) and 93% for ColoSense. See the colorectal cancer chapter.
The blood algorithms are not screening tests, and no protein tumor marker is recommended for general population cancer screening. Their numbers also shift between studies. OVA1 specificity was 54% in its pivotal study and 42% in a 2026 payer review, which still judged the evidence insufficient to show better health outcomes. PHI and the 4Kscore help with biopsy decisions after a raised PSA. In its pivotal data the 4Kscore had 76.4% sensitivity and 70.3% specificity. More in the ovarian and prostate chapters.
In studies
Protein plus DNA and AI-protein panels in studies
No multi-cancer blood test is FDA approved as of 1 October 2026. Some multi-marker versions are sold as laboratory-developed tests (LDTs), usually without FDA review. Others remain in studies. The emerging technologies chapter covers their regulatory status.
- Cancerguard (Exact Sciences): an LDT launched on 10 September 2025 that measures tumor DNA and protein levels. Sensitivity was 64.1% in its development study and 55.6% in its validation study, both at 97.4% specificity and both case-control. Its performance in real screening is not yet known.
- OncoSeek (SeekIn): seven protein tumor markers plus clinical data and an AI algorithm. A 2025 study of 15,122 participants on four immunoassay platforms reported 58.4% sensitivity and 92.0% specificity. Sensitivity rose from 42.8% in stage I to 79.7% in stage IV, and results varied between cohorts (area under the curve 0.744 to 0.912). The authors call the design predominantly case-control. The company page states no regulatory status.
Two cautions follow. Case-control studies compare known patients with people without cancer, which tends to flatter a test. At 92% specificity, the worked example above gives a PPV near 4% in a low-prevalence population. See our guide to AI multi-biomarker blood tests.
Methods
Research methods and reporting standards
Biomarker research follows five phases, set out by Pepe and colleagues in 2001.
The five phases of biomarker development, with checks that apply throughout. Adapted by OncoFirm™ from Pepe et al., J Natl Cancer Inst 2001, and the US Institute of Medicine (2012).
The PRoBE design (prospective specimen collection, retrospective blinded evaluation) requires that people are enrolled before diagnosis and that samples from cases and controls are collected and processed the same way. Such studies are large: at an incidence of about 0.5%, 20,000 people are needed to obtain 100 colon cancers.
- TRIPOD+AI (2024): 27 main items for studies that develop or evaluate prediction models, including machine learning. It replaces TRIPOD 2015.
- STARD-AI (2025): 40 items for diagnostic accuracy studies of AI tests, with attention to datasets, bias and fairness.
- REMARK: reporting recommendations for tumor marker prognostic studies.
Three checks separate a trustworthy panel from a promising one. Lock-down: the algorithm and threshold are fixed before clinical evaluation. External validation: testing uses specimens collected independently, ideally at another institution. Calibration: predicted risks match observed risks. Overfitted models, which have learned noise in their training data, give risk estimates that are too extreme. Assay differences between settings also harm calibration, so an algorithm trained on laboratory analyzer values cannot be assumed to work on another device.
Programmes
Large research programmes
- EDRN: the US National Cancer Institute's Early Detection Research Network, which uses the five-phase pathway and the PRoBE design.
- CSRN Vanguard Study: a feasibility study of up to 24,000 people, launched in 2025, to prepare a much larger randomized trial of multi-cancer tests.
- UK Biobank proteomics: in 41,931 people and about 3,000 plasma proteins, signatures of only 5 to 20 proteins beat basic clinical models for 67 of 218 diseases. The authors say the findings need validation in external studies.
- ACED: the International Alliance for Cancer Early Detection, founded in 2019 by Cancer Research UK with university partners in the UK and US. Dana-Farber and DKFZ joined in February 2025.
Lessons
Lessons from failures
- A valid marker without benefit: UKCTOCS followed more than 200,000 women for about 16 years. Screening based on CA-125 found 39% more stage I and II cancers and 10% fewer stage III and IV cancers, yet it did not reduce deaths from ovarian cancer.
- Early proteomic signatures: outside researchers showed that an early serum protein pattern test for ovarian cancer rested on artifacts in the data, not on biology.
- Overfitting and unlocked models: a 2012 US Institute of Medicine report described tests used in clinical trials before their models were locked down. It warned that reusing training samples for validation overstates accuracy.
The NHS-Galleri trial missed its primary endpoint in 2026. According to the National Cancer Institute, the highest level of evidence is a fall in deaths in a randomized controlled trial. See the early detection evidence hub.
Point of care
Practical limits of multiplex lateral flow
Measuring several markers on one lateral flow strip adds analytical problems to the statistical ones.
- Line count: a 2025 review found that more than three to four test lines per strip harms readability and needs wider membranes, raising production cost by up to 40%.
- Cross-reactivity: antibodies for one target can bind another. The review cites cross-reaction in about 30% of tested multiplex strips in one infectious disease example.
- Dynamic range: markers differ in concentration range and units, and each line can show a hook effect (falsely low readings at very high levels).
- Algorithm transfer: each line's imprecision feeds into the combined score. The algorithm must be trained, locked and validated on values from the reader that will be used.
In the sources reviewed, quantitative multiplex lateral flow for cancer proteins exists only as proof of concept. We found no cleared multi-marker cancer algorithm that runs on a lateral flow platform. See the fluorescent lateral flow platform guide and tumor-associated antigens.
OncoFirm™ roadmap
OncoFirm™ MultiDx: a research concept
OncoFirm™ MultiDx is a research concept for measuring more than one protein marker on a fluorescent lateral flow platform. It is for research use only. It carries no screening or multi-cancer claims, and it is not cleared or approved by the FDA or any other regulatory authority.
Single markers first
The CEA (designed range 1–100 ng/mL) and PSA (0.5–50 ng/mL) assays are in development, with planned traceability to WHO international standards (CEA 73/601; PSA 2nd IS 17/100). See the CEA guide.
One reader, one strip
The one-reader, one-strip design is meant to support a growing menu of tests. The result-time target of 20 minutes or less is a development target.
Validation before claims
Any panel would need analytical validation of each line, a locked algorithm, external validation and calibration on reader values. Clinical collaborations are the route to that evidence.
Pipeline
AFP is a future pipeline candidate (see the AFP guide). TF antigen assay concepts are concept-stage. Nothing is for sale.
OncoFirm™ assays, reader and software are in development, have not been cleared or approved by the FDA or any other regulatory authority, are for research use only and are not for sale. Designed ranges and result times are development targets, not validated performance claims.
FAQ
Frequently asked questions
What is multi-biomarker analysis?
It means combining two or more biomarkers, often with age or other clinical data, in one algorithm that gives a single score. The aim is better accuracy than any single marker. The gain only counts if it holds up in independent validation.
Why does specificity matter so much in cancer screening?
Most people who are screened do not have cancer, so even a small false-positive rate creates many false alarms. In our worked example (0.5% prevalence, 60% sensitivity), positive predictive value is about 3.6% at 92% specificity and 37.6% at 99.5%. This is arithmetic, not a study result.
Are ROMA, OVA1, PHI and the 4Kscore cancer screening tests?
No. ROMA, OVA1 and Overa are for women who already have a pelvic mass and are being assessed before surgery. PHI and the 4Kscore help with biopsy decisions in men with a raised PSA.
What are TRIPOD+AI and STARD-AI?
They are reporting checklists. TRIPOD+AI (2024) has 27 main items for prediction model studies, including machine learning. STARD-AI (2025) has 40 items for diagnostic accuracy studies of AI tests.
Is OncoFirm™ MultiDx a multi-cancer detection test?
No. OncoFirm™ MultiDx is a research concept for quantitative multi-marker measurement, for research use only. It carries no screening or multi-cancer claims, is not cleared or approved by the FDA or any other regulatory authority, and is not for sale.
Cancer Atlas 2026
More in the Atlas
Related OncoFirm™ guides
Sources
References
- National Cancer Institute. Cancer Screening Overview (PDQ), Health Professional Version. Updated 16 October 2023. www.cancer.gov/about-cancer/screening/hp-screening-overview-pdq
- NCI Early Detection Research Network. Five-Phase Approach and PRoBE Study Design. edrn.cancer.gov/about-edrn/five-phase-approach-and-prospective-specimen-collection-retrospective-blinded-evaluation-study-design/
- Pepe MS, et al. Five phases of biomarker development (EDRN-hosted copy). J Natl Cancer Inst. 2001;93:1054–1061. edrn.cancer.gov/documents/156/FivePhasesofBiomarkerDevelopment.pdf
- The ASCO Post. Study Shows Blood Test Can Identify Multiple Cancers in Asymptomatic Women (Lennon et al., Science 2020). 10 June 2020. ascopost.com/issues/june-10-2020/study-shows-blood-test-can-identify-multiple-cancers-in-asymptomatic-women/
- ADLM Clinical Laboratory News. Multianalyte Assays With Algorithmic Analysis in Women's Health. July 2018. myadlm.org/cln/articles/2018/july/multianalyte-assays-with-algorithmic-analysis-in-womens-health
- Fujirebio Diagnostics. FDA clearance announcement for the ROMA test (HE4 and CA 125). Press release, September 2011. www.fujirebio.com/en-us/news-events/fda-clears-nextgeneration-biomarker-test-to-determine-likelihood-of-ovarian-cancer-in
- FEP Blue. Medical Policy 2.04.62: multimarker serum testing related to ovarian cancer. Effective 1 April 2026. www.fepblue.org/-/media/PDFs/Medical-Policies/2026/March/Pharmacy-Policies/New-Policies/20462-Multimarker-Serum-Testing-Related.pdf
- U.S. Food and Drug Administration. Premarket Approval P090026: Access Hybritech p2PSA (Prostate Health Index). 14 June 2012. www.accessdata.fda.gov/scripts/cdrh/cfdocs/cfpma/pma.cfm?id=P090026
- U.S. Food and Drug Administration. Summary of Safety and Effectiveness Data, P190022: 4Kscore Test. 7 December 2021. www.accessdata.fda.gov/cdrh_docs/pdf19/P190022B.pdf
- Exact Sciences. FDA Approves Exact Sciences' Cologuard Plus Test. Press release via BioSpace, October 2024. www.biospace.com/press-releases/fda-approves-exact-sciences-cologuard-plus-test-setting-a-new-benchmark-in-non-invasive-colorectal-cancer-screening
- Geneoscopy. FDA Approves ColoSense, a Multi-Target Stool RNA Colorectal Cancer Screening Test. Press release, May 2024. www.geneoscopy.com/fda-approves-colosense-geneoscopys-noninvasive-multi-target-stool-rna-mtrna-colorectal-cancer-screening-test/
- Exact Sciences. Launch of the Cancerguard multi-cancer early detection blood test. Press release, 10 September 2025. www.exactsciences.com/news-events/press-releases/exact-sciences-launches-cancerguard-first-of-its-kind-multi-cancer-early-detection-blood-test
- Lustgarten Foundation. From CancerSEEK to Cancerguard. September 2025. lustgarten.org/from-cancerseek-to-cancerguard-how-early-research-sparked-a-new-era-in-cancer-detection/
- Shen et al. Multi-centre study of OncoSeek (seven protein tumor markers plus AI). npj Precision Oncology. 2025;9:321. www.nature.com/articles/s41698-025-01105-2
- Collins GS, Moons KGM, Dhiman P, et al. TRIPOD+AI statement. BMJ. 2024;385:e078378. www.bmj.com/content/385/bmj-2023-078378
- Sounderajah et al. STARD-AI reporting guideline. Nature Medicine. 2025;31:3283–3289. www.nature.com/articles/s41591-025-03953-8
- EQUATOR Network. REMARK: Reporting Recommendations for Tumour Marker Prognostic Studies (McShane et al., Br J Cancer 2005;93:387–391). www.equator-network.org/reporting-guidelines/reporting-recommendations-for-tumour-marker-prognostic-studies-remark/
- Van Calster et al. Calibration of clinical prediction models. BMC Medicine. 2019;17:230. link.springer.com/article/10.1186/s12916-019-1466-7
- Institute of Medicine. Evolution of Translational Omics, chapter 6. National Academies Press; 2012. www.nationalacademies.org/read/13297/chapter/8
- University College London. Screening for ovarian cancer did not reduce deaths (UKCTOCS, The Lancet 2021). May 2021. www.ucl.ac.uk/news/2021/may/screening-ovarian-cancer-did-not-reduce-deaths
- Velotta et al. Review of multiplexed lateral flow assays. Sensors. 2025;25(17):5414. www.mdpi.com/1424-8220/25/17/5414
- Carrasco-Zanini et al. Sparse plasma protein signatures for disease prediction in UK Biobank. Nature Medicine. 2024. www.nature.com/articles/s41591-024-03142-z
- National Cancer Institute. Q&A About the Cancer Screening Research Network (CSRN). prevention.cancer.gov/research-areas/networks-consortia-programs/csrn/q-a-about-csrn
- Early Cancer Institute, University of Cambridge. Two new partners join the International Alliance for Cancer Early Detection. 20 February 2025. www.earlycancer.cam.ac.uk/news/two-new-partners-join-international-alliance-cancer-early-detection-aced
About this article. Compiled from regulator documents, reporting guidelines, peer-reviewed papers and company releases; several performance figures come from manufacturers, and the PPV table and seven-marker example are our own arithmetic. Part of the OncoFirm™ Cancer Atlas 2026, written by the OncoFirm™ Scientific Team from the sources linked above and reflecting public information as of 1 October 2026. It is for education and is not medical advice. OncoFirm™ has no affiliation with the companies or tests named. OncoFirm™ assays are in development, have not been cleared or approved by the FDA and are not available for sale.
Collaborate on the next edition
Clinicians, laboratories, researchers and industry partners can contribute data, corrections or validation studies. Collaboration inquiry · Investor inquiry