Oncology data

Last reviewed: 2026-06-28 — see the freshness policy.

Learning objectives

After this chapter you will be able to:

• Explain why oncology data is uniquely complex and multi-modal.
• Name the core oncology standards —ICD-O-3, AJCC TNM staging, mCODE, NAACCR.
• Design a data flow that unifies pathology, genomics, staging, treatment, and outcomes.

Why oncology is its own data problem

Cancer care produces some of the most complex longitudinal data in medicine. A single patient's record spans pathology (histology, grade), staging (how advanced), genomics (somatic mutations driving therapy), imaging, treatment lines (surgery, radiation, systemic therapy), and response/outcomes over years. It is also reporting-mandated — providers must report cancers to central registries — and fast-evolving, as precision oncology ties new biomarkers to new drugs constantly. Generic clinical data models capture only part of this; oncology needs its own standards.

flowchart LR
  Path["Pathology<br/>(histology, grade)"] --> Model
  Stage["Staging<br/>(AJCC TNM)"] --> Model
  Gen["Genomics<br/>(somatic variants)"] --> Model
  Tx["Treatment lines<br/>+ response"] --> Model
  Model["Unified oncology model<br/>(mCODE / OMOP)"] --> Reg["Registry reporting"]
  Model --> Trial["Clinical-trial matching"]
  Model --> Board["Tumor board / decision support"]
  Model --> RWE["RWE / outcomes research"]

The core standards

mCODE is the one to anchor on: it standardizes the primary cancer condition, secondary/metastatic disease, TNM staging, tumor markers, cancer-related genomics, treatments, and disease status as FHIR resources — so oncology data can move between EHRs, registries, trial-matching services, and research without custom interfaces.

Precision oncology: genomics drives therapy

Modern oncology is genomics-led. An NGS panel (or whole-genome/exome sequencing) on a tumor produces somatic variants, which are interpreted (OncoKB/ClinVar, tiered by AMP/ASCO/CAP guidelines) to select targeted therapies. The architecture must therefore link the variant store to the clinical record:

• Distinguish somatic (tumor-acquired) from germline (inherited) variants — different consent, interpretation, and reporting.
• Carry the variant → interpretation → therapy chain so a tumor board (and an audit) can see why a therapy was chosen.
• mCODE's genomics profiles (built on the FHIR Genomics work) are the standard way to represent this in the record.

Architecture implications

• Unify multi-modal data. Pathology, staging, genomics, imaging, and treatment arrive from different systems; mCODE (FHIR) is the conformance target, OMOP the analytics model (with an oncology extension). Map source codes carefully — histology inICD-O-3, diagnoses in ICD-10/SNOMED (terminologies).
• Registry reporting is a pipeline, not an afterthought. NAACCR reporting is mandated; design the extract early. mCODE-to-registry tooling (CodeX) reduces manual abstraction.
• Clinical-trial matching is a flagship oncology use case: structured mCODE data (stage, biomarkers, prior therapy) matches patients to trial eligibility — a strong AI/agent opportunity.
• Staging is hard. TNM has clinical vs pathologic variants, edition dependencies, and site-specific rules; treat it as a first-class, validated data element, not a free-text field.

Pain points & best practices

• Histology/site coding (ICD-O-3) and staging require trained abstraction — expect data-quality work and validate against registry rules.
• Biomarkers move fast — the variant-to-therapy knowledge base changes constantly; version it (like vocabularies) and date interpretations.
• Somatic vs germline governance — germline findings are inherited PHI for the family; handle consent and return-of-results deliberately (see de-identification & consent).
• Anchor on mCODE for interoperability; don't invent a local oncology model.

Lab

Extend the hls-fhir-interop lab with mCODE profiles, and join tumor variants from hls-variant-store-aws to the clinical record. Generate synthetic oncology patients with the hls-synthea-data lab (Synthea has cancer modules).

Check yourself

1. Why isICD-O-3needed alongside ICD-10/SNOMED for cancer data?
2. What does mCODE standardize, and why anchor an oncology platform on it?
3. Why must somatic and germline variants be governed differently?

Further reading

• mCODE Implementation Guide · CodeX FHIR Accelerator
• AJCC TNM staging · NAACCR
• ICD-O-3 (WHO) · SEER · OncoKB