Ask a labeling team which version of the prescribing information is currently approved, and you will usually get an answer quickly. Ask how they know, and the answer is often less comfortable: it is the latest one in the folder, or the latest one filed under Module 1.14, or the one somebody marked as final.

None of those is the same as “approved”. The latest filing in 1.14 may be a draft with tracked changes, a negotiation round, or a later supplement’s proposal that has not been acted on yet. Treating it as the approved label is how an outdated or never-approved text ends up as the baseline for the next change.

Approval is an event, not a file

A label becomes approved because the agency took an approval action. So the only defensible way to say which version was approved is to tie it to that action: the last version filed before the approval, where the approval itself comes from the agency’s record — the approval letter on file, or the action recorded in Drugs@FDA.

That is how Label History in DnXT Reviewer works. For each labeling role — prescribing information, patient labeling, carton and container, SPL — it lists every version filed under 1.14 in filing order, with a status: proposed, each negotiation round, approved, final labeling, a later proposed change, removed. Agency labeling requests filed alongside a version are linked to it.

Two rules keep it honest:

  • Approval is never inferred from the label itself, and never from being the latest filing.
  • An approval window with no labeling draft is not credited. A CMC supplement approved with no labeling change does not make the last label “newly approved”.

The familiar “Current Approved” label disappears. In its place is “Latest filed”, which is true, and a status for each version that says how it was established.

Comparing the proposal with the approved text

The next question is always the same: what did the agency change? Between the proposal that opened an approval window and the label that came out of it, which sections moved, and how?

Most comparison tools answer that with a visual diff of two PDFs. That is where accuracy quietly breaks. PDF text extraction loses structure, splits words across lines, reorders columns and drops formatting that carries meaning. A diff built on it produces both false changes and missed ones — and in labeling, a missed change can mean a market label that no longer matches its approval.

Exact sources only

DnXT’s label comparison follows a short list of rules, because an inaccurate comparison is worse than none:

  • Only exact sources are compared: structured product labeling (SPL) XML, or a clean Word document with only accepted text. PDFs are never compared.
  • Like for like only. SPL against SPL, Word against Word. A mixed pair is refused, with the reason.
  • Only whitespace is normalised. Case, punctuation and numbers all count.
  • Every section must rebuild exactly. If the changes reported for a section cannot reconstruct both versions precisely, the section is shown as “not compared” rather than shown with a guess.
  • No silent choices. If a sequence holds more than one candidate prescribing information, the comparison refuses rather than picking one.

Comparison runs section by section, so the reviewer sees that, for example, the indication was narrowed, two pregnancy and lactation subsections were rewritten and a clinical studies section was folded into another — not a wall of red and green.

The default pairing is the one teams most often need: the proposal that opened the most recent approval window, against the first label published on DailyMed after that approval. Any two filed versions can be chosen instead.

Evidence behind each section

Where annotated labeling was filed, each section of the prescribing information can also show the CTD documents the sponsor cited for it. Those citations are read from the annotation hyperlinks themselves, so a reviewer can open the clinical pharmacology summary behind section 2.2 or the specification behind the description with one click — and a citation pointing at a document that is not in the filing is flagged.

Why it matters

Labeling is where regulatory precision meets commercial consequence. Every market label, every CCDS comparison and every future change starts from an assumption about what was approved. Making that assumption provable — and making the comparison exact — removes a quiet source of risk from every labeling decision that follows.

DnXT builds eCTD publishing, submission planning, document management and dossier review software for regulatory operations teams. Book a demo to see Label History and exact submitted-vs-approved comparison on your own filed labeling.