NDC to ATC Therapeutic Classification Crosswalk

Description

Map any NDC to its WHO therapeutic classification in one flat file. No API wrangling required.

The Problem: A Simple Question That Takes Days to Answer

You have a list of National Drug Codes (NDCs) and you need to know: what therapeutic classes are these drugs in? It sounds simple. It is not.

NDC codes identify drug products — specific packages from specific manufacturers. They tell you nothing about what the drug actually does. To get therapeutic classifications, you need the WHO Anatomical Therapeutic Chemical (ATC) system, the global standard for drug utilization research.

Getting from NDC to ATC means chaining multiple government APIs: NDC to RxNorm concept, then RxNorm to ATC. You need to handle rate limits (20 requests/second across 300,000+ codes), deduplicate RxNorm concepts, resolve multi-ingredient ambiguities, and build the full ATC hierarchy from level-4 codes up to level-1 organ systems.

Most analysts give up partway through and use incomplete workarounds.

The Solution: One Flat File, Complete Hierarchy

We ran the full pipeline so you don’t have to. Every active NDC in the FDA directory has been resolved through NLM’s RxNorm API to its ATC level-4 chemical subgroup, with the complete hierarchy (levels 1 through 4) derived and included in every row.

• ATC Level 1: Anatomical main group (e.g., N = Nervous system)
• ATC Level 2: Therapeutic subgroup (e.g., N02 = Analgesics)
• ATC Level 3: Pharmacological subgroup (e.g., N02B = Other analgesics and antipyretics)
• ATC Level 4: Chemical subgroup (e.g., N02BE = Anilides)

NDCs that map to multiple ATC-4 codes get multiple rows — this is a true many-to-many crosswalk, not a lossy one-to-one approximation.

Why level 4, not level 5? ATC level 5 (the individual chemical substance code, e.g., A10BX16 for tirzepatide) is part of the WHO’s licensed classification system and is not available through NLM’s public RxNorm API. RxNorm’s ATCPROD and ATC sources top out at level 4 — the chemical subgroup. Level 4 is the most granular therapeutic classification available from public US government data, and it is the standard used in most drug utilization research.

Technical Specifications

• Grain: One row per (ndc_code, atc4_code) pair. This is a many-to-many crosswalk — an NDC can belong to multiple ATC-4 therapeutic subgroups (e.g., combination drugs), and each ATC-4 code covers many NDCs.
• Format: Flat .csv (Comma Separated Values), UTF-8 encoded
• Columns: ndc_code, product_name, rxcui, atc4_code, atc4_name, atc3_code, atc3_name, atc2_code, atc2_name, atc1_code, atc1_name
• NDC Formatting: String type with hyphens and leading zeros preserved exactly as FDA publishes them
• Coverage: ~60-76% of active FDA NDCs (medical devices, veterinary, and some OTC products lack ATC mappings — this is an inherent limitation of the ATC system, not a data quality issue)
• Provenance: Every row includes the RxCUI intermediate code for full traceability back to NLM’s source data
• Delivery: Instant digital download (ZIP containing CSV and metadata.json with row counts, SHA-256 checksum, and coverage statistics)

Perfect For

• Health Services Researchers: Drug utilization studies, pharmacoepidemiology, claims data analysis
• Pharmacy Benefit Managers: Formulary classification, therapeutic substitution analysis, drug spend analytics
• Pharma Data Engineers: Building drug master data, enriching claims pipelines, standardizing across NDC and ATC systems
• Health Economists: MEPS-style analysis, cost-effectiveness research by therapeutic class

The Business Case

Commercial drug classification databases cost $500+ per year. Building this yourself means writing rate-limited API code, running it for 5+ hours, debugging edge cases in multi-ingredient drugs, and rebuilding whenever the FDA updates its directory.

For $49, you get the complete crosswalk file, ready to join against your data in minutes.