Companion diagnostics sit at the interface of medical-device and therapeutic regulation. The test supplies information essential to the safe and effective use of a corresponding drug or biological product, so regulators examine not only analytical performance but also the connection between the assay, the clinical evidence, and the therapeutic labeling.
This guide outlines major regulatory considerations for CDx programs, including intended use, coordinated development, investigational testing, analytical and clinical validation, manufacturing, software, bridging, and lifecycle changes. Requirements depend on jurisdiction, product, risk, and development path; teams should verify current agency guidance for their program.
The intended-use statement anchors the device. It identifies the analyte or biomarker, specimen, test method or system, patient population, corresponding therapy, clinical purpose, and any limitations. Every major study should support this claim. A validation performed with a different specimen or algorithm may not establish performance for the proposed use.
Distinguish companion diagnostics from prognostic, monitoring, screening, and complementary tests. A biomarker can have several roles, but the regulatory claim depends on how the result is used. Early alignment across protocol, statistical analysis, diagnostic development plan, drug strategy, and proposed labels prevents small wording differences from becoming major evidence gaps.
Therapeutic and diagnostic teams should coordinate agency interactions, milestones, data cutoffs, and responses. Regulators may need to understand how the test selected or stratified trial participants, how assay failures were handled, and whether the commercial test matches the clinical-trial version.
In the United States, FDA’s 2014 final guidance describes the general expectation that an IVD companion diagnostic and therapeutic product be approved or cleared contemporaneously, except in limited circumstances. Current plans should be confirmed with the relevant review divisions.
| Domain | Representative questions | Typical evidence |
|---|---|---|
| Analytical validity | Does the test measure the biomarker reliably? | Accuracy, precision, sensitivity, specificity, cutoff, interference, stability, robustness |
| Clinical validity/performance | Does the result identify the intended therapy population? | Trial data, clinical bridging, outcome association, prespecified analyses |
| Specimen validity | Does performance apply to the claimed specimen? | Collection, handling, stability, tissue sufficiency, matrix studies |
| Software and algorithm | Is result generation controlled and reproducible? | Verification, validation, locked rules, cybersecurity and data integrity as applicable |
| Manufacturing | Can the test and critical reagents be produced consistently? | Process controls, specifications, lot release, stability, supplier controls |
| Labeling | Can users apply and interpret the result correctly? | Intended use, warnings, limitations, workflow, result interpretation |
When a diagnostic result determines trial eligibility, treatment assignment, or other clinical management, applicable investigational-device requirements must be assessed. Risk determination, informed consent, institutional review, monitoring, recordkeeping, and labeling responsibilities depend on the study and jurisdiction. The testing laboratory also needs procedures for sample accountability, training, quality control, invalid results, deviations, and result reporting.
Assay version control is critical. Clinical data should be traceable to the exact reagent lots, software version, cutoff, and workflow used. Changes during enrollment need documented assessment. A research assay may evolve, but teams must preserve enough evidence and specimens to bridge to the final CDx.
Challenge variant classes, concentrations, near-neighbors, wild-type background, and the complete claimed range.
Measure repeatability and reproducibility near the clinical boundary across lots, sites, instruments, operators, and days.
Evaluate endogenous and exogenous interferents, specimen quality, timing, temperatures, and controlled workflow variation.
Validation design should reflect technology. NGS adds coverage, bioinformatics, sequence context, and variant-class validation; immunohistochemistry adds fixation, staining, reader, and scoring variables. Internal links to precision, linearity, and recovery evaluation and interference and matrix-effect evaluation describe relevant assay studies.
The strongest evidence uses the finalized test prospectively in the therapeutic trial. In practice, prototype or trial assays may differ from the commercial CDx. Bridging can compare analytical agreement and evaluate the clinical consequences of discordant results using retained specimens and outcome data. The design depends on the changes, biomarker prevalence, sample availability, and test performance.
Bridging is not merely a correlation plot. It should account for missing specimens, selection bias, invalid results, and uncertainty around the cutoff. Changes in specimen type, analyte, platform, or algorithm may require more evidence than a reagent-lot update. Planning sample consent, storage, and statistical methods early protects options later.
Enzymes, antibodies, conjugates, substrates, primers, controls, calibrators, and software can all be critical. Specifications should link raw-material attributes to assay risks. For enzymes this can include identity, purity, concentration, activity, contaminating nucleases or proteases, formulation, stability, and functional testing in the final reaction.
Supplier qualification, change notification, incoming acceptance, lot bridging, traceability, storage, shipping, and backup supply belong in the control strategy. Batch-to-batch consistency validation and assay transfer and manufacturability assessment help connect reagent controls with commercial production.
Many molecular CDx devices include signal processing, quality rules, databases, and variant-classification algorithms. Requirements, architecture, verification, validation, access control, audit trails, cybersecurity, and version management should be proportionate to the software’s role. Training data and thresholds should be controlled when statistical or machine-learning methods are used.
After authorization, changes to enzymes, suppliers, manufacturing sites, instruments, panels, software, cutoffs, or labels require documented assessment and potentially regulatory submission. A change that appears minor at raw-material level may alter limit of detection or patient classification. Lifecycle plans should include complaint trending, nonconformance, corrective action, postmarket obligations, and expansion to new therapies or biomarker groups.
Regulatory pathways differ across regions. In the European Union, companion diagnostics fall within the IVDR framework and involve interaction between notified-body conformity assessment and consultation with a medicines authority or competent authority, as applicable. Timelines and evidence packages should be planned with regional experts.
In oncology, FDA has guidance on developing and labeling certain IVD companion diagnostics for a specific group of oncology therapeutic products. Group labeling is evidence-based, not automatic class extrapolation. FDA also operates a voluntary pilot concerning performance information for certain oncology drug products used with specific IVDs; the program does not change applicable approval standards. Teams should consult current official materials rather than relying on a static summary.
CDx technical documentation and regulatory support can help organize this evidence, while final regulatory decisions remain with the appropriate authorities.
Whatever platform or specimen is selected, development should begin with a written link between intended use and analytical requirements. Define the patient population, biomarker, specimen, treatment decision, reportable result, turnaround expectation, and use environment. Then identify the failure modes that could change classification: target loss, nonspecific signal, amplification bias, reagent drift, interference, software error, or an invalid result that delays therapy. This risk map determines which enzyme attributes and assay controls deserve the most attention.
Feasibility experiments should include representative clinical material as early as possible. Purified templates and synthetic controls are valuable for isolating variables, but they do not reproduce fixation damage, low tumor fraction, endogenous inhibitors, sample heterogeneity, or extraction carryover. A staged study can begin with controlled materials, add individual challenges, and then confirm performance in specimens spanning the intended range. Samples near the cutoff are especially informative because small shifts in recovery, background, or signal can change the treatment category.
Critical enzymes should be specified by more than catalog activity. Identity, purity, concentration, specific activity, contaminating nuclease or protease limits, formulation, storage, and functional performance may all be relevant. The release method should use conditions that predict performance in the diagnostic reaction. When the vendor activity assay and CDx chemistry differ substantially, an assay-level incoming or bridging test can provide a more direct control. Multiple lots should be evaluated before pivotal use so the acceptance range reflects manufacturing variation rather than one favored batch.
Robustness studies intentionally vary parameters that will move in practice: reaction time and temperature, pipetting, sample input, operator, instrument, reagent lot, shipping excursion, and storage duration. Interference studies should use justified concentrations and combinations of endogenous substances, collection additives, medications, and process residuals. Controls must fail when the vulnerable step fails; an abundant control target may remain positive even when a low-copy clinical target is lost.
Finally, document changes across the full measurement system. A new enzyme lot, buffer, primer pool, conjugation process, extraction kit, instrument, or software version can alter analytical performance even if the intended use is unchanged. Risk-based comparability should focus on the attributes most likely to affect the cutoff and claimed range. Preserving retained samples, reference materials, version history, and a predefined bridging strategy makes lifecycle improvements possible without breaking the connection to the clinical evidence.
These questions keep development centered on the treatment decision rather than isolated technical metrics. They also create a common language for biomarker, clinical, regulatory, quality, manufacturing, and supplier teams.
A CDx regulatory strategy succeeds when the test’s intended use, analytical performance, clinical evidence, manufacturing controls, and therapeutic labeling tell one coherent story. Early coordination and disciplined version control reduce bridging risk. Because requirements and agency programs evolve, developers should use official current guidance and obtain program-specific regulatory advice.
Creative Enzymes supports diagnostic enzyme selection, engineering, formulation, conjugation, analytical evaluation, and scale-up for molecular and immunoassay CDx workflows.
