MedDRA is the common language of safety data. Every case, every line listing and every signal analysis depends on coding that is consistent from one processor to the next, and on a clear policy for what happens when the dictionary changes.
How the dictionary is built
MedDRA is a standardised medical terminology maintained under the auspices of ICH, and it is released twice a year, in March and September. It is organised in five levels, from the very specific Lowest Level Terms (LLTs) through Preferred Terms (PTs) and High Level Terms and High Level Group Terms to System Organ Classes (SOCs). The PT is the level at which most analysis is done, while the LLT is the level at which coders usually select, because it is closest to what the reporter actually wrote.
Principles of good term selection
The ICH points-to-consider documents on term selection and data quality are the best practical reference. In outline:
- Code what was reported, not what you think the reporter meant. If the wording is ambiguous, query it.
- Do not code a term that is not in the report, and do not add diagnoses the reporter did not make.
- Where a diagnosis is reported together with its signs and symptoms, follow a documented convention rather than leaving the choice to the individual.
- Code each distinct medical concept, and avoid combining unrelated events into one term.
- Code the indication and the medical history as carefully as the adverse event; they are used in analysis too.
Versions, recoding and legacy data
A new release changes terms: some are moved, some are made non-current, some are added. Companies need a written policy on which version is used for new cases and when the safety database is upgraded. Without one, a signal search run across several years of data can give a misleading answer simply because the same event was coded differently in different releases. Version management belongs in the validation and change control of the safety database, not in the head of one specialist.
Auto-coding and its limits
Most safety systems offer automatic coding, and it is a sensible way to reach a consistent first answer for common terms. It is not a substitute for review. Misspellings, local-language terms, abbreviations and multi-concept phrases are where automated coding goes wrong. A reasonable model is to accept auto-coding for exact matches and to route everything else to a human coder, with periodic reconciliation of a sample of both.
Searching with MedDRA
Standardised MedDRA Queries (SMQs) group terms relating to a defined medical condition, and are widely used in signal detection and aggregate reports. They only work if the underlying coding is sound.
How PQRA helps
PQRA can review your coding conventions, version management policy and quality checks, and help write or update the procedures that sit behind them. Where coding is outsourced, we can help define what you should expect from the vendor and how to verify it.
Contact PQRA to discuss your pharmacovigilance system, in Greece and across the EU.


No responses yet