Ask a publisher what fails most often in final validation and the list is rarely dramatic. Fast Web View is off. Bookmarks change the reader’s zoom. The author’s name is still in the document properties. A file type sits in a section where the agency does not accept it. A hyperlink opens the right document on the wrong page.

None of these is hard to fix. All of them are expensive to find at the end, when every correction means a new rendition, a new checksum and another validation run against a deadline.

The answer is to move the checks — and the safe fixes — to where documents are created and approved.

Check documents where they start

DnXT’s document QC checks Word and PDF documents as they are uploaded and versioned in the eDMS. Administrators decide which checks run, at what severity and with what limits, by company and region — rather than every customer living with one fixed list.

Each company also decides what a finding means. In inform mode, findings are shown and nothing is blocked. In block mode, a document cannot be approved while it has a high-severity finding open — or while QC has not finished, because “QC did not run” is never the same as “QC passed”. Users with the right permission can override with a reason, and that reason is written to the document’s audit trail.

Fix the mechanical problems automatically

Some findings have exactly one correct fix. DnXT applies those and checks the result in as a new version:

  • opening view set to the first page, with the bookmarks panel when there are bookmarks
  • bookmarks collapsed and set to inherit zoom
  • author metadata cleared — in both places PDFs store it
  • Fast Web View applied

Every fix is re-checked with the same rule that found the problem. A fix that does not make the check pass is reported as skipped — never as fixed. Font embedding is deliberately not offered: a font that is missing from a file cannot be added without the original font, and pretending otherwise would hide a real defect.

FDA forms are treated differently, for a sensible reason. A form is submitted exactly as FDA publishes it, so the checks about how the file was built are skipped for forms, while the checks that describe the filled-in form — size, structure, page size, links, blank pages — still run.

Fast Web View without breaking the checksum

Optimising a PDF for fast web view rewrites the file — and a rewritten file no longer matches the checksum the sequence recorded for it. DnXT applies it at the moment a file is placed, before the checksum is taken, so the sequence always describes the exact file that ships. Published sequences are never touched.

Links that land where they say

Hyperlinks fail in a quieter way. A link can be technically valid — the target file exists, the page exists — and still be wrong, because page 41 is not where section 2.7.3 starts.

DnXT’s Auto Link Verification checks every link once for each version of a PDF, and keeps the result:

  • the target is part of this submission, or of an earlier sequence of the same application
  • the destination page exists
  • the landing page actually contains what the link text claims — a section number, a table or figure, a study ID, or the heading words — and, if not, which page does

These checks are fixed rules, not a model’s opinion. Each one comes back passed, failed or not checkable, so a reviewer starts from evidence. It applies to every link, whoever or whatever created it.

Gates at publishing time

Two final checks close the loop. A sequence cannot move to publishing while any document placed from the eDMS is still a draft; the refusal names each document and its status. And for US submissions, every file is checked against FDA’s published table of accepted file formats by eCTD location — held as data, so an FDA revision is a data update that names its version.

Re-validation then answers the question publishers actually have: did my changes add anything? Validation history compares each run with the previous one and lists what is new, what was fixed and what is still failing.

The result

None of this is glamorous. It is the work of making sure the defects that are easy to fix are fixed early, the fixes never create a new problem, and every link a reviewer clicks goes where it promises. That is what lets final publishing be about the submission, not the PDFs.

DnXT builds eCTD publishing, submission planning, document management and dossier review software for regulatory operations teams. Book a demo to see document QC, auto-fix and Auto Link Verification on a document set of your own.