Coding with MedDRA: Conventions, Term Selection, and SMQs

MedDRA’s size is also its risk: with roughly 90,000 lowest level terms to choose from, two equally competent coders can reasonably land on different LLTs for the same verbatim unless they’re working from the same written rules. Coding conventions exist to close that gap — documented, organisation-specific principles for handling misspellings, abbreviations, combination terms, "due to" concepts, and "always query" terms like chest pain, so that coded data stays consistent both within a company and across the wider regulatory ecosystem that exchanges it.

The ICH-endorsed foundation for those conventions is the MedDRA Term Selection: Points to Consider (MTS:PTC), a living document refreshed every March and recommended as the basis for every organisation’s own coding rules. Its general principles run through the whole coding workflow: quality of source data comes first, since no amount of careful coding fixes a badly collected verbatim; quality assurance means qualified individuals review term selection and provide human oversight of any autoencoded results; MedDRA’s structure itself is never altered ad hoc; only current LLTs are selected, chosen for maximum specificity rather than a vague match; and — critically — coders select terms for everything reported, including device issues, product quality issues, and medical history, without adding information the source never contained.

That last principle is where the discipline really shows. If a case reports abdominal pain plus elevated serum amylase and lipase, the correct approach is three separate LLTs, not a single inferred LLT for "pancreatitis" — even though any clinician reading the case would suspect exactly that. The MTS:PTC’s Section 3 works through dozens of these judgment calls in detail: diagnoses with or without signs and symptoms, death and other patient outcomes, suicide and self-harm, combination terms, age vs. event specificity, medication errors and misuse, drug interactions, off-label use, and product quality issues among them. The common thread is always the same — code what was said, as specifically as MedDRA allows, and use medical judgment only to match an existing term, never to manufacture one MedDRA doesn’t already contain.

Standardised MedDRA Queries build directly on well-coded data. An SMQ is a validated grouping of PTs — and every LLT subordinate to those PTs — assembled around a medical concept like hepatic disorders or anaphylactic reaction, so a safety database can be queried consistently rather than reinvented case by case. In clinical trials, where a product’s safety profile isn’t yet established, multiple SMQs are typically run routinely as a screening tool; post-marketing, selected SMQs support signal detection, single-case alerts, and periodic aggregate reporting. None of it works, though, if the underlying LLT selection was inconsistent — which is exactly why the signal-detection and feedback-loop lesson back in Module 3 rests on the coding discipline covered here.