Since 1 April 2026, every new application to Japan’s PMDA is built in eCTD 4.0 and checked against several hundred rules that PMDA publishes as configuration. Most are straightforward. A handful decide whether your first sequence is accepted.
Key takeaways
- PMDA publishes its 4.0 rules as data — which element, which check, which parameters, and when the rule is in force.
- The checks cover five layers of a submission unit, from the package on disk to the characters in a document title.
- Three traps catch most teams: conditions written only in the rule text, retired rules, and validators that only check their own output.
- Ask any vendor how they handle all three before your first Japan 4.0 sequence.
A rulebook published as data
Most regulators publish validation criteria as a document: a numbered list of rules, each with a severity. PMDA goes further for eCTD 4.0. Its rules come as configuration tables that state, for each check, the element it applies to, the function that performs it, the parameters it uses and whether the rule is currently switched on.
That is good news for anyone building or evaluating a publishing tool, because rules published as data can be enforced as data rather than re-interpreted by a developer. It also means the details matter. A tool that reads the tables loosely will pass sequences PMDA rejects — or reject sequences PMDA would accept.
What PMDA checks
At the base is the package itself: the folders, file names and sizes, the checksum that anchors the submission, and agreement between the sequence number and its folder. Above that sits the submission message, where PMDA checks that the required elements are present — the receiving authority, the submission-unit identifiers and codes, the application identity and the category event — and that a long list of elements is absent.
That second half is easy to overlook. A tool that never writes a forbidden element will never fail the rule on its own output, so the rule is never tested until a package arrives from somewhere else.
The next layers are about values and text: fixed values that must be exact, the code systems in use and codes that must exist in them, length limits, and PMDA’s permitted character set for brand and generic names, the applicant name, document titles and keyword values. Study data carries its own file tags and naming rules on top.
A validator that only checks its own output will never find the problems in packages it did not build.
Where teams get caught
Conditions in the text. Some requirements apply only “at first-version submission” or “in the case of” something, and that condition appears only in the description, not in the structured rule. Applied literally, two rules contradict each other, and either your first sequence or every later one fails.
Retired rules. PMDA switches rules off with an effective date. A validator that ignores the date can fail you on a rule that no longer applies — and send your team chasing a problem that does not exist.
Only your own output. The packages most likely to carry hidden defects are the ones you did not build: from a partner, a CRO or a previous vendor. If your validator cannot check them, you inherit their problems at the gateway.
Before your first Japan 4.0 sequence, ask any vendor three questions: How do you apply conditions that exist only in the rule text? How do you treat rules PMDA has retired? Will your validator check a package your tool did not produce?
How DnXT approaches it
DnXT publishes and validates Japan eCTD 4.0 against PMDA’s configuration as published. The configuration files are bundled as PMDA issues them, so an update is a file replacement rather than a code change. Rules PMDA has switched off are excluded by their effective window, conditions stated only in the rule text are applied so first and later sequences are judged correctly, and the same checks run on imported and externally built packages.
See Japan eCTD 4.0 on your own content
Bring a package — yours or a partner’s — and we will run PMDA’s rules on it with you.