Precision oncology rests on a foundational premise: that a treatment's benefit is not uniformly distributed across a patient population, but concentrated in a subset defined by a molecular, protein, or genetic characteristic that can be measured before treatment begins. The instrument that measures this characteristic — and determines whether a given patient should receive a given drug — is the companion diagnostic. As of 2025, the FDA has approved or cleared more than 50 companion diagnostic devices for use in oncology, spanning tissue-based immunohistochemistry, fluorescence in situ hybridisation, PCR mutation assays, next-generation sequencing panels, and liquid biopsy platforms.
This article approaches companion diagnostics at the depth the topic demands. We examine the regulatory architecture governing CDx development and approval, the analytical and clinical validation science that underpins that approval, the co-development timeline and strategic decisions that determine whether a CDx reaches approval alongside its paired therapeutic, and four landmark case studies that illustrate the spectrum of CDx complexity in current oncology practice.
Defining Companion Diagnostics: The Regulatory Threshold
The FDA defines a companion diagnostic as "an in vitro diagnostic (IVD) device or imaging tool that provides information essential for the safe and effective use of a corresponding therapeutic product." The critical word in this definition is essential. A CDx is not a helpful add-on to prescribing — it is a regulatory prerequisite. The drug's prescribing information (label) specifies that the CDx test must be performed and a positive result obtained before the drug is administered to a patient.
This is distinct from a complementary diagnostic, which provides information that is useful but not mandatory. A complementary diagnostic might help select patients who are more likely to respond, but the label does not require a positive test result as a condition of prescribing. The regulatory pathway for each is different, and the design of clinical trials intended to generate registration-enabling data must account for this distinction from the earliest stages of drug development.
Companion Diagnostic (CDx)
Essential for drug use. The drug's FDA-approved label requires a positive CDx result before prescribing. Regulated as a Class III medical device under 21 CFR Part 809. Requires Premarket Approval (PMA) — FDA's most rigorous device approval pathway. Must be submitted and reviewed simultaneously with the drug NDA/BLA.
Complementary Diagnostic
Provides useful but non-essential information. The drug may be used without a positive result. May use 510(k) or De Novo regulatory pathway depending on risk classification. Drug label may reference the diagnostic without requiring its use as a condition of prescribing. Lower regulatory burden but also lower reimbursement leverage.
FDA Regulatory Framework: PMA Under 21 CFR Part 809
Companion diagnostics are regulated by FDA's Center for Devices and Radiological Health (CDRH) as in vitro diagnostics under 21 CFR Part 809. Because a CDx is deemed essential for safe drug use, it is classified as a Class III device — the highest risk classification under the Medical Device Amendments of 1976 — and must obtain Premarket Approval (PMA) before it can be marketed.
The PMA Pathway
The PMA application must demonstrate reasonable assurance of safety and effectiveness for the specific intended use. For a CDx, this means demonstrating that the device accurately identifies the patient population in whom the drug is safe and effective. The PMA dossier for a CDx typically contains:
- Analytical performance data: sensitivity, specificity, accuracy, precision (repeatability and reproducibility), specimen stability studies, and interference studies
- Device description and manufacturing information
- Software documentation (if applicable — increasingly relevant for NGS-based CDx with bioinformatics pipelines)
- Clinical data demonstrating that patients who test positive on the CDx benefit from the corresponding therapeutic (and, critically, that patients who test negative do not benefit — or benefit substantially less)
- Proposed labelling specifying the CDx's intended use, limitations, and expected performance in the clinical setting
Global Regulatory Landscape
CDx regulation outside the US is harmonised in principle but operationally fragmented. The European Medical Devices Regulation (EU MDR 2017/745) and the companion In Vitro Diagnostic Regulation (EU IVDR 2017/746) came into full effect in 2022 and significantly raised CDx regulatory standards across Europe. Japan's PMDA and China's NMPA have developed their own CDx frameworks, often with different clinical evidence requirements than FDA. For global oncology trials, this means the biomarker assay strategy must account for multiple regulatory destinations simultaneously — a CDx approved by FDA may not have an equivalent approval in the EU or Japan at the time of drug approval.
Analytical and Clinical Validation: The Science Behind CDx Approval
Analytical Validation
Analytical validation establishes that the CDx device measures the biomarker with acceptable accuracy and reproducibility. The FDA expects analytical validation studies to be conducted according to CLSI (Clinical and Laboratory Standards Institute) guidance documents, particularly:
| Analytical Parameter | Definition | Oncology CDx Relevance |
|---|---|---|
| Sensitivity (Analytical) | Lowest concentration or copy number the device can reliably detect | Critical for FISH assays (HER2 amplification at low copy number ratio), liquid biopsy ctDNA (allele frequency as low as 0.1%), MSI testing in microsatellite-low tumours |
| Diagnostic Sensitivity (Clinical) | Proportion of truly biomarker-positive specimens correctly classified as positive | A CDx with 90% clinical sensitivity will misclassify 10% of eligible patients as test-negative — those patients miss treatment. FDA scrutinises sensitivity closely for enrichment CDx |
| Diagnostic Specificity | Proportion of truly biomarker-negative specimens correctly classified as negative | Low specificity leads to false-positive results — biomarker-negative patients receive the drug without evidence of benefit, increasing toxicity exposure without clinical gain |
| Precision: Repeatability | Reproducibility of results when the same specimen is tested multiple times by the same operator, same instrument, same day | Within-run variation must be below the analytical threshold that would change patient classification (e.g., above/below the IHC 3+ scoring threshold) |
| Precision: Reproducibility | Reproducibility across different operators, instruments, laboratory sites, and days | Especially important for IHC-based CDx (subjective scoring) and NGS-based CDx (bioinformatics pipeline variation). FDA expects multi-site, multi-operator reproducibility data |
Clinical Validation: The Biomarker-by-Treatment Interaction
Clinical validation is the most scientifically demanding aspect of CDx development. It requires demonstrating that biomarker status — as measured by the specific CDx device — predicts differential treatment benefit. This is not the same as showing that biomarker-positive patients do better overall (that would be a prognostic biomarker). The CDx must demonstrate a statistically significant biomarker-by-treatment interaction: the magnitude of treatment effect must be substantially larger in biomarker-positive patients than in biomarker-negative patients.
The design and statistical power required for this interaction test have major implications for clinical trial sample size. A trial powered only on the biomarker-positive population (enrichment design) cannot formally test the interaction. A trial that enrolls biomarker-positive and biomarker-negative patients in the same randomised design can test the interaction, but requires a substantially larger sample size to have adequate power for both the main effect and the interaction test simultaneously.
CDx Co-Development: Timeline and Strategic Decisions
The ideal CDx co-development programme runs in parallel with the therapeutic development programme from early Phase I — not as an afterthought when the drug reaches Phase III. Industry experience consistently shows that late-stage CDx development is the most common cause of drug-CDx approval misalignment, where the drug is approved but no CDx is available, or the drug is delayed because the CDx is not ready for co-submission.
| Development Phase | Therapeutic Activities | CDx Activities |
|---|---|---|
| Pre-IND / Early Phase I | Identify molecular target; assess biomarker hypothesis from preclinical data; IND filing | Select candidate biomarker assay; initiate assay development; assess commercial CDx platform options; begin pre-IND biomarker strategy discussion with FDA |
| Phase I Dose Escalation | Safety and PK; dose escalation; expansion cohort biomarker enrichment if early signals emerge | Evaluate candidate assay performance on Phase I archival specimens; identify CDx partner if in-house development is not planned; begin analytical development and feasibility testing |
| Phase I/II / Phase II | Initial efficacy signal; identify biomarker-defining patient population; proof-of-concept in enriched population | Lock prototype CDx assay for Phase II; complete analytical performance package; initiate clinical bridging studies if laboratory assay ≠ commercial assay; begin pre-submission meetings with CDRH |
| Phase III (Pivotal) | Randomised confirmatory trial in biomarker-defined population using CDx as inclusion criterion | Prospective CDx testing of all enrolled patients using the commercial-intent assay; accumulate PMA analytical and clinical data; submit PMA in parallel with NDA/BLA |
| Submission / Review | NDA/BLA submission to CDER; labelling negotiations including CDx requirement in prescribing information | PMA review by CDRH; joint CDER-CDRH communication; advisory committee if warranted; CDx approval synchronised with drug approval |
Landmark Case Studies
The Original CDx: HercepTest and Herceptin (1998–2024)
The HER2/trastuzumab (Herceptin) pairing is the archetype of companion diagnostic development — and a cautionary tale about the consequences of CDx complexity. Trastuzumab was approved in 1998 for HER2-positive metastatic breast cancer, with the HercepTest (IHC, Dako) as the first CDx ever approved by FDA. The treatment produced dramatic responses: in the pivotal Slamon et al. trial (2001, NEJM), trastuzumab + chemotherapy improved OS from 20.3 months to 25.1 months (p=0.046) in patients with IHC 3+ or FISH-amplified HER2.
What followed was two decades of diagnostic complexity that arguably delayed optimal patient selection. Four different HER2 testing platforms (IHC 3+, FISH amplification, FISH ratio ≥2.0, ISH) were used in different trials and different countries, with different cutoffs. A landmark 2007 ASCO/CAP guideline reanalysis found that up to 20% of previously classified "HER2-positive" patients may have been incorrectly classified — some receiving trastuzumab without benefit, others denied it when they might have benefited. The 2023 ASCO/CAP HER2 guideline update added a HER2-low category (IHC 1+, or IHC 2+ with ISH not amplified), establishing a new patient population for trastuzumab deruxtecan (T-DXd) benefit that was invisible to the original binary IHC 3+/FISH framework.
Sequential CDx Development: From First-Generation to Third-Generation EGFR TKI
The EGFR-mutant NSCLC story illustrates how CDx requirements evolve with therapeutic generations. When erlotinib was approved for NSCLC in 2013, the cobas EGFR Mutation Test v1 was approved simultaneously as the CDx, detecting exon 19 deletions and L858R mutations in tumour tissue. First- and second-generation EGFR TKIs (erlotinib, gefitinib, afatinib) produced impressive initial responses — median PFS approximately 9–13 months — but all patients eventually developed resistance, the most common mechanism being the T790M resistance mutation, occurring in ~50–60% of first-line progressors.
Third-generation osimertinib (Tagrisso) was designed to overcome T790M resistance. The cobas EGFR Mutation Test v2 was approved in 2015 as a CDx for T790M testing in plasma (liquid biopsy) for patients progressing on prior EGFR TKI — the first plasma-based companion diagnostic approved by FDA. The FLAURA trial (2018) then established osimertinib as first-line standard of care in EGFR-mutant NSCLC with a dramatically superior PFS benefit (18.9 vs. 10.2 months, HR 0.46), and the cobas EGFR test was subsequently approved for first-line EGFR mutation detection in both tissue and plasma. The FoundationOne CDx simultaneously holds approval as a CDx for osimertinib in tissue.
Multi-Tumour, Multi-Drug CDx Complexity: BRACAnalysis CDx and FoundationOne CDx
The BRCA1/2 story demonstrates CDx complexity when a single biomarker drives drug approvals across multiple tumour types and multiple drugs. Olaparib (Lynparza) was first approved in 2014 for germline BRCA1/2-mutated ovarian cancer, with the BRACAnalysis CDx (Myriad Genetic Laboratories, germline whole blood testing) as the companion diagnostic — the first CDx based on germline genetic testing rather than tumour testing. Subsequent approvals expanded BRACAnalysis CDx to include breast cancer (2018), pancreatic cancer (2019), and metastatic castration-resistant prostate cancer (2020), all for olaparib.
Simultaneously, somatic BRCA testing (tumour tissue) became clinically relevant because ~5–7% of patients have somatic BRCA mutations without germline alterations. FoundationOne CDx added BRCA1/2 somatic testing as part of its pan-tumour NGS panel, providing an alternative CDx pathway that detects both somatic and germline-equivalent alterations in tumour tissue. In prostate cancer, the CDx landscape expanded further: olaparib in mCRPC uses FoundationOne CDx as a CDx for a broader set of HRR genes (14 genes including BRCA1/2, ATM, BARD1, CDK12, and others) — not just BRCA1/2. This created a situation where the CDx for the same drug in the same tumour type could be either germline BRACAnalysis or tumour-based FoundationOne CDx, depending on the specific approved indication.
The Continuous Biomarker Challenge: PD-L1 IHC and TMB
Pembrolizumab (Keytruda) has the most complex companion diagnostic profile in oncology — because PD-L1 expression is a continuous variable with context-dependent clinical significance that varies by tumour type, line of therapy, and pembrolizumab indication. The PD-L1 IHC 22C3 pharmDx (Dako/Agilent) is the primary approved CDx, but the required threshold differs dramatically across indications. For first-line NSCLC monotherapy, a tumour proportion score (TPS) ≥50% is required — only approximately 30% of NSCLC patients qualify. For first-line combination with chemotherapy, TPS ≥1% is required. For cervical cancer second-line monotherapy, TPS ≥1% is the cutoff. In other indications — MSI-H/dMMR tumours and TMB-high solid tumours — PD-L1 testing is not required at all; the relevant CDx is either a validated MMR IHC assay or FoundationOne CDx (TMB ≥10 mut/Mb, approved in June 2020 across all solid tumours, PMA P170019/S016).
The TMB threshold approval — the first tumour-agnostic biomarker-defined indication using a quantitative genomic parameter — represented a conceptual advance over prior CDx frameworks: instead of identifying presence or absence of a mutation, FoundationOne CDx here identifies an aggregate genomic property (total somatic mutation burden per megabase of sequenced genome) that predicts immunotherapy response across multiple tumour types regardless of histology.
NGS Panel CDx and the FoundationOne CDx Paradigm
The approval of FoundationOne CDx in 2017 was a watershed moment in companion diagnostic regulation. Prior CDx approvals had been single-analyte devices — one assay measuring one biomarker for one drug. FoundationOne CDx is a broad NGS panel that simultaneously detects alterations across 324 cancer-related genes, reports TMB and MSI status, and has been approved as a CDx for multiple drugs across multiple tumour types using a single tissue sample.
This creates what is effectively a CDx bundle: a single test result can determine eligibility for multiple drugs simultaneously. A patient with advanced NSCLC tested by FoundationOne CDx might simultaneously receive results relevant to osimertinib (EGFR mutation), alectinib (ALK fusion), dabrafenib/trametinib (BRAF V600E), capmatinib (MET exon 14 skipping), pembrolizumab (TMB-high), and olaparib (BRCA2 somatic mutation) — all from a single formalin-fixed paraffin-embedded (FFPE) tissue block.
Liquid Biopsy as CDx: ctDNA Platforms and Current Limitations
Liquid biopsy — the detection of circulating tumour DNA (ctDNA) or circulating tumour cells (CTCs) from a blood sample — offers the prospect of CDx testing without requiring tissue biopsy. The only FDA-approved liquid biopsy CDx as of 2025 is Guardant360 CDx (Guardant Health), approved for the detection of EGFR exon 19 deletions and exon 21 L858R mutations in plasma for first-line osimertinib treatment of NSCLC. An additional plasma CDx approval exists for cobas EGFR Mutation Test v2 (plasma) for osimertinib in the T790M setting.
Advantages of ctDNA CDx
Minimally invasive (blood draw, not tissue biopsy). Can capture tumour heterogeneity from multiple sites simultaneously. Repeatable — can track mutation evolution during treatment. Particularly valuable when tissue is insufficient, inaccessible, or biopsy is high risk.
Limitations of ctDNA CDx
Lower sensitivity than tissue in early-stage disease (low tumour shedding). False-negative rate up to 30% in stage I–II NSCLC. Some structural rearrangements (fusions, amplifications) have poor sensitivity in plasma DNA. A negative plasma result does not exclude the mutation — tissue biopsy must follow. Not yet approved for most biomarkers beyond EGFR in NSCLC.
In the clinical trial context, protocols using ctDNA CDx should pre-specify the algorithm for patients with negative plasma results — typically mandating tissue CDx testing as a reflex second step, with randomisation eligibility determined by either positive plasma or positive tissue result. This plasma-first, tissue-reflex approach maximises patient access while maintaining regulatory alignment with approved CDx labelling.
Implications for Clinical Trial Design and Operations
The CDx strategy is not a regulatory afterthought — it shapes nearly every aspect of how a biomarker-enriched oncology trial is designed and operated.
CDx Readiness Checklist for Clinical Trial Teams
CDx Market and Emerging Frontiers
The companion diagnostic market reflects the pace of precision oncology drug development. The global CDx market was valued at approximately $8–10 billion in 2024, with compound annual growth projections of 18–22% through 2030, driven by the expanding catalogue of biomarker-defined oncology approvals and the increasing adoption of NGS-based CGP for routine clinical care in major cancer centres.
Several emerging frontiers will further expand the CDx landscape in the next 5–10 years. Multi-omic CDx — integrating genomic, transcriptomic, proteomic, and epigenomic biomarkers into a single predictive model — is moving from research to early regulatory evaluation, with FDA's Artificial Intelligence/Machine Learning Device Action Plan providing an emerging framework for multi-feature algorithm-based CDx. Minimal residual disease (MRD) detection as a CDx endpoint for adjuvant therapy decisions is under active regulatory development — FDA published draft guidance on MRD as a regulatory endpoint in haematologic malignancies in 2020, with solid tumour applications in development. Spatial transcriptomics and digital pathology AI-powered image analysis CDx are in pre-submission discussions with FDA, with the potential to extract biomarker information from standard H&E slides without additional staining or molecular testing.
Frequently Asked Questions
Related education on kclgmedical.com: Education Hub · Protocol Preparation in Oncology Trials · Biospecimen Handling in Oncology Trials · ICH GCP E6(R3) Guide