Indications

An indication is a specific approved use of a drug for a disease — the FDA label entry that says “pembrolizumab is indicated for the first-line treatment of metastatic NSCLC in patients whose tumors express PD-L1 (TPS >= 1%).” Four entities decompose that sentence into queryable structure: indications, therapeutic_approaches, indication_biomarkers, and therapeutic_approach_prior_therapies.

Domain context

In pharma, a single drug approval can cover multiple indications. Drug X may be approved for disease Y in first-line treatment with biomarker requirement W, and separately for disease Z in second-line after prior platinum therapy. Each of those is a distinct indication. The indications domain captures this granularity: the disease, the treatment context (line, setting, stage), the biomarker requirements, and the prior therapy prerequisites.

Entities

erDiagram
    drug_approvals ||--o{ indications : "indicated for"
    indications }o--|| diseases : "treats"
    indications ||--o{ indication_biomarkers : "requires"
    indication_biomarkers }o--|| biomarkers : "biomarker"
    indication_biomarkers }o--o| therapeutic_approaches : "scoped to"
    indications ||--o{ therapeutic_approaches : "treatment strategies"
    therapeutic_approaches }o--o| regimens : "uses"
    therapeutic_approaches ||--o{ therapeutic_approach_prior_therapies : "requires prior"
    therapeutic_approach_prior_therapies }o--o| drugs : "prior drug"

indications — approved use linking drug_approval to disease

The central table. Each row links one drug_approval to one disease with rich contextual facets.

Column	Type	Purpose
drug_approval_id	FK	Parent drug approval
disease_id	FK	Target disease
indication_type	text	initial, supplemental, or expanded
indication_text	text	Original indication wording from the drug label
disease_stages	jsonb	Applicable stages (e.g., ["Stage IIIB", "Stage IV"])
disease_subtypes	jsonb	Applicable subtypes (e.g., ["squamous", "non-squamous"])
disease_extents	jsonb	Disease extents (e.g., ["metastatic", "locally advanced"])
disease_statuses	jsonb	Patient statuses (e.g., ["treatment-naive", "relapsed"])
population	text	Age or fitness constraints (e.g., "adult", "pediatric >= 12")
risks	jsonb	Boxed warning risk tags
accelerated_approval_date	timestamptz	Date of accelerated approval, if applicable
full_approval_date	timestamptz	Date of full/traditional approval, if applicable
accelerated_approval_studies	jsonb	NCT IDs or study names supporting accelerated approval
full_approval_studies	jsonb	NCT IDs or study names supporting full approval
data_source	text	Provenance: bioloupe, fda, ema, llm, etc.
source_reference	text	External ID or URL in the source system
confidence	decimal(3,2)	Extraction confidence score (0.00—1.00), LLM-extracted only

A check constraint ensures full_approval_date >= accelerated_approval_date when both are set — an indication cannot receive full approval before accelerated approval.

therapeutic_approaches — treatment strategies within an indication

Each indication can have multiple therapeutic approaches representing different treatment strategies (monotherapy vs. combination, different lines of therapy).

Column	Type	Purpose
indication_id	FK	Parent indication
regimen_id	FK (nullable)	Treatment regimen used, if mapped
is_single_agent	boolean	Monotherapy (true) vs. combination (false)
treatment_lines	jsonb	Lines of therapy (e.g., ["1L", "2L"])
treatment_settings	jsonb	Treatment settings (e.g., ["neoadjuvant", "adjuvant"])
min_line	integer	Minimum line of therapy (numeric)
max_line	integer	Maximum line of therapy (numeric)

indication_biomarkers — biomarker requirements for patient selection

Captures companion diagnostic requirements. A biomarker can be scoped to the indication as a whole or to a specific therapeutic approach within it.

Column	Type	Purpose
indication_id	FK	Parent indication
therapeutic_approach_id	FK (nullable)	Scoped to a specific approach, or null for indication-wide
biomarker_id	FK	The required biomarker
requirement	text	required, recommended, or optional
required_value	text	Expected test result (e.g., "TPS >= 50%", "Positive")
companion_diagnostic	text	Name of the approved companion diagnostic test

therapeutic_approach_prior_therapies — required prior treatments

Records which prior therapies a patient must have received (or failed) before qualifying for a therapeutic approach.

Column	Type	Purpose
therapeutic_approach_id	FK	Parent therapeutic approach
therapy_name	text	Name of the required prior therapy
requirement	text	must_have_received, refractory_to, relapsed, or intolerant_to
drug_id	FK (nullable)	Linked drug entity, when the prior therapy maps to a known drug

Design decisions

Indications hang off drug_approvals, not drugs. A drug may have approvals in multiple regions (FDA, EMA, PMDA, NMPA), each with different indications. Attaching indications to the approval rather than the drug preserves this regional specificity. Querying “all indications for drug X” is a simple join through drug_approvals.

Therapeutic approaches model the treatment landscape. Rather than flattening line-of-therapy and setting into columns on indications, the schema uses a separate therapeutic_approaches table. This handles the common case where a single indication covers multiple treatment contexts — e.g., pembrolizumab for NSCLC is approved as both monotherapy in 1L and in combination with chemotherapy in 1L, each a distinct therapeutic approach.

Biomarkers can scope to indication or approach. The nullable therapeutic_approach_id on indication_biomarkers supports both indication-wide biomarker requirements (e.g., “all uses of this drug require HER2 testing”) and approach-specific ones (e.g., “PD-L1 >= 50% required only for the monotherapy approach”).

Prior therapies reference drugs loosely. The drug_id on therapeutic_approach_prior_therapies is nullable because prior therapy requirements often reference drug classes or regimen names (e.g., “prior platinum-based chemotherapy”) that do not map cleanly to a single drug entity.

Example queries

All indications for a drug, with disease and approval region:

select
  i.indication_text,
  d.name as disease,
  da.region,
  i.accelerated_approval_date,
  i.full_approval_date
from indications i
join drug_approvals da on da.id = i.drug_approval_id
join diseases d on d.id = i.disease_id
where da.drug_id = :drug_id
order by i.full_approval_date desc nulls last;

Indications requiring a specific biomarker with their companion diagnostics:

select
  i.indication_text,
  d.name as disease,
  ib.required_value,
  ib.requirement,
  ib.companion_diagnostic
from indication_biomarkers ib
join indications i on i.id = ib.indication_id
join diseases d on d.id = i.disease_id
join biomarkers b on b.id = ib.biomarker_id
where b.name = 'PD-L1'
  and ib.requirement = 'required';

Second-line therapeutic approaches with their prior therapy requirements:

select
  dr.name as drug,
  d.name as disease,
  ta.treatment_lines,
  ta.is_single_agent,
  tapt.therapy_name,
  tapt.requirement as prior_therapy_requirement
from therapeutic_approaches ta
join indications i on i.id = ta.indication_id
join drug_approvals da on da.id = i.drug_approval_id
join drugs dr on dr.id = da.drug_id
join diseases d on d.id = i.disease_id
left join therapeutic_approach_prior_therapies tapt on tapt.therapeutic_approach_id = ta.id
where ta.treatment_lines @> '["2L"]'::jsonb
order by dr.name, d.name;