Clinical development plan
Clinical development: building evidence before making claims
This is the plan for taking a research-stage cancer biomarker assay from analytical work to clinical feasibility. It is written for the investigators, sites, laboratories and statisticians who would carry it out with us.
The principle
Develop the technology. Generate the data. Validate the performance. Then establish the claim. Nothing in this plan has been completed yet, and the plan itself will change as partners and regulators weigh in.
The clinical question comes first. A biomarker assay is only useful if a defined population, at a defined point in care, would be managed differently because of its result. Intended use therefore drives every design choice that follows, from specimen type to comparator.
Nine stages, each with an exit criterion
Stages overlap in practice, but none is skipped.
1. Clinical need and intended use
Define the population, the setting, the decision the result informs and the comparator. Exit: a written intended-use statement reviewed by clinical advisors.
2. Biomarker selection
Choose markers with published evidence and a plausible clinical role. Exit: documented rationale for each analyte.
3. Assay development
Antibody pairs, reporter chemistry, membrane and cassette design. Exit: a design-frozen research prototype.
4. Reader and software verification
Verify that the reader and its software measure and report as specified. Exit: verification report and locked software version.
5. Analytical performance
Limit of detection, precision, linearity, interference, cross-reactivity, hook effect, stability, lot-to-lot. Exit: analytical report against pre-set criteria.
6. Clinical feasibility
Retrospective studies on banked specimens with known status. Exit: preliminary agreement with reference methods and a go/no-go decision.
7. Clinical performance
Prospective studies in the intended-use population, with independent sites. Exit: sensitivity, specificity and predictive values with confidence intervals.
8. Usability and human factors
Show that intended users can run the test correctly in the intended setting. Exit: human-factors summary.
9. Regulatory submission planning
Pathway, predicate or classification questions and evidence gaps, raised with regulators early. Exit: an evidence plan revised after regulator feedback.
Study design commitments
- Human-subjects protection. All specimen and participant research runs under IRB review or a documented exemption, with informed consent as the protocol requires.
- Prespecified analysis. Hypotheses, endpoints, sample size and acceptance criteria are fixed with the study statistician before data are unblinded.
- Representative populations. Cases and controls reflect the intended-use population, including the benign and inflammatory conditions that cause false positives.
- Independent data. Any model is trained, tuned and tested on separate datasets, with a final test set that developers never see. The reasoning is set out in our article on validating AI in cancer diagnostics.
- Reference standards. Results are compared with accepted reference methods and clinical truth, defined in the protocol.
- More than one site. Performance is confirmed outside the developing laboratory before any claim is made.
Regulatory pathway
The pathway depends on the final intended use and on FDA classification. Possible routes for an in vitro diagnostic include premarket notification (510(k)), De Novo classification and premarket approval (PMA). We expect to seek FDA feedback through a pre-submission before pivotal studies are designed.
Naming a possible pathway does not mean FDA has reviewed, cleared, classified, approved or endorsed any OncoFirm product. Quality-system and risk-management work begins during development so that validation lots are made under controlled conditions. See the manufacturing and regulatory priorities.
Roles for clinical and research partners
Clinical work is where partners matter most. Clinical and academic research collaboration describes how these relationships run day to day.
Principal investigators
Lead or co-lead a feasibility study, shape the intended-use statement and co-author results.
Hospitals and cancer centers
Serve as a study site, contribute consented specimens and bring the clinical context a protocol needs.
Clinical laboratories
Run reference-method testing and method-comparison studies under your quality system.
Biobanks
Provide annotated specimens with documented collection, processing and storage history.
Biostatisticians
Own the statistical analysis plan, sample-size justification and independent validation of any model.
Contract research organizations
Manage multi-site studies, monitoring and data management.
Regulatory and quality consultants
Advise on pathway, design and development controls and submission strategy.
Frequently asked questions
Has any clinical study been run yet?
No. Clinical feasibility work has not started.
Which specimen types are planned?
Candidate specimen types include whole blood, plasma and serum. The choice will be made per assay and confirmed in analytical studies.
Will results be published?
That is the intention. The plan is to publish analytical and clinical results in peer-reviewed journals, with partner investigators as authors under standard authorship criteria.
Can a site join a study that another institution leads?
Yes. Multi-site designs are expected at the clinical performance stage, and additional sites strengthen a proposal.
Help design the first clinical feasibility study
If you lead clinical research in oncology, laboratory medicine or biostatistics, we would like to hear how you would approach it.
Development and regulatory notice
OncoFirm™ technologies described on this site are in research and development. Their performance characteristics have not been established, and no OncoFirm product has been cleared, approved or authorized by the U.S. Food and Drug Administration. Nothing here is medical advice. Reference to a federal agency or funding program does not imply endorsement, funding, sponsorship or affiliation.
AI-generated or software-generated information is not intended to substitute for the judgment of qualified healthcare professionals.