FDA is running two webinars on 23 September for an audience it defines narrowly: people "who develop labeling for prescription drugs regulated under NDAs or biological products regulated under BLAs," along with clinical pharmacology and safety specialists. One session covers the DRUG INTERACTIONS section of the prescribing information and the other covers QTc information. Of all the documents a development programme produces, the label is among the most consequential, and among the few for which FDA rarely publishes instruction on the writing itself. The QTc session rests on a separate guidance, finalised in December 2025, and is not the subject here. Behind the drug interaction session sits a draft guidance from October 2024, Drug Interaction Information in Human Prescription Drug and Biological Product Labeling, which is unusual among FDA guidances in how far down into the prose it goes.
Most labeling guidance concerns what information belongs where, whereas this one goes further and specifies word choice, sentence order and voice.
The preferred phrasings are named
The guidance asks that instructions to prevent or manage a clinically significant interaction "must be accurate, specific, and practical and should be actionable for the HCP." It then does something labeling guidances do less often at this level of granularity, which is to name the words:
Applicants should use specific statements rather than vague or ambiguous statements (e.g., "avoid concomitant use" is preferred over "use with caution," "reduce dosage" is preferred over "adjust dosage") and use active voice instead of passive voice.
That is editorial direction of a kind normally found in a house style guide rather than in a document carrying regulatory weight. "Use with caution" and "adjust dosage" appear throughout approved labels, and the guidance is saying that both are too vague to be useful to a prescriber. Precision of that kind is what the rest of the submission is already judged on, from a cited protocol synopsis to the figures a label ultimately carries.
The accompanying example makes the standard concrete, because for an interaction with warfarin, a recommendation describing "a specific change in the frequency or timing of international normalized ratio (INR) monitoring is more informative for HCPs than a non-specific recommendation such as 'monitor INR' that reflects standard clinical practice for patients taking warfarin." A sentence that tells a clinician to do what they already do is treated as a sentence that says nothing.
The order of the sentences is specified
The guidance sets an order for describing an interaction, and gives the reason for it. Descriptions "should generally contain information presented in the following order to promote consistency, as applicable: (1) instructions for preventing or managing the clinically significant DIs, (2) mechanisms of the clinically significant DIs, and (3) clinical effects." The stated rationale is that "this labeling approach prioritizes the information that is most actionable."
This inverts how these paragraphs are often drafted, since the familiar pattern opens with the pharmacology, establishes the mechanism, describes the exposure change, and arrives at the recommendation last. The guidance asks for the recommendation first, on the reasoning that the reader is a prescriber making a decision rather than a reviewer following an argument. The audience is defined explicitly as health care providers "including those with limited clinical pharmacology expertise." That is a narrower and more demanding reader than the reviewer most sections are drafted for, and it is a further reason the label is contested territory, as the skinny-label litigation has made clear for labeling teams.
A ladder of model sentences
The guidance supplies drafted sentences for each level of severity, which function as a controlled vocabulary. Where the risk clearly outweighs any possible benefit, the section states that concomitant use "is contraindicated." Where use is inadvisable without rising to a contraindication, it states "Avoid concomitant use of DRUG-X with drugofen-a."
Where use is generally unavoidable, because the interacting drug treats a serious or life-threatening condition, the guidance asks for actionable measures alongside the avoidance, and drafts that too: "Avoid concomitant use of DRUG-X with drugofen-a. If concomitant use is unavoidable, obtain ECGs prior to initiating, during concomitant use, and additionally as clinically indicated." A fourth rung covers modifying dosage or administration, where concomitant use necessitates a change rather than avoidance.
Having model sentences at four severity levels turns part of the drafting judgement into a structured selection, since the clinical judgement about which rung applies still belongs to the sponsor. That is a meaningful change for a writer, and it is the same move as reviewers publishing the lens they apply rather than leaving sponsors to infer it. It also matters for any tooling that reads labels, whether on the reviewer’s side or the sponsor’s, because consistent phrasing is what makes labels comparable across products, and it lands hardest on new modalities writing their first labels.
Structure, subsections and headings
The guidance is equally specific about architecture, and pharmacokinetic interactions typically go into one of two subsections, 7.1 "Effects of Other Drugs on DRUG-X" or 7.2 "Effects of DRUG-X on Other Drugs," and a bidirectional interaction goes in both "in order to accurately describe both DIs." Pharmacodynamic interactions generally get their own subsections titled for the interacting drug or class.
Where an interaction combines mechanisms, the heading itself has to carry them, and the guidance gives the form: "Combined P-gp and Moderate CYP3A Inhibitors," extended for a population as "Combined P-gp and Moderate CYP3A Inhibitors in Patients with Renal Impairment." There is also a note that any "floating content" sitting between the section 7 heading and subsection 7.1 should be moved into the appropriate subsection.
On format, complex or extensive information, which the guidance quantifies as three or more interactions, "could be more effectively conveyed in a table rather than in text." Where a table is used, a sentence preceding it should describe what it contains, and the guidance drafts that sentence as well.
What the section is not for
Two exclusions matter for anyone deciding what to cut. The section "should not describe DIs that are not clinically significant," unless communicating the absence of one is itself important. And "in vitro and/or animal data alone should not be included in this section because they are generally insufficient to justify the presence of a clinically significant DI," with a recommendation to consult FDA where a sponsor believes otherwise.
The section, or required content within it, is omitted entirely where clearly inapplicable. Cross-references are the mechanism for avoiding repetition where interaction information appears in more than one place, which is the sort of thing that only works if the cross-references exist while the document is being drafted rather than being added at the end.
Interactions serious enough to appear in CONTRAINDICATIONS or WARNINGS AND PRECAUTIONS must additionally be discussed in more detail in the DRUG INTERACTIONS section, so the same interaction is written more than once, at different depths, for different purposes.
What this changes in the drafting
The guidance is a draft, and it is nonbinding, so none of this is a requirement in the ordinary sense. What it represents is a review division writing down, at sentence level, what it has been asking for in review cycles. Where a sponsor disagrees, the route runs through the review process rather than a single forum, whether in labeling negotiations, written responses to information requests, or a formal meeting, and the package carrying that last kind of question has its own constraints on how many questions there is room for.
For a labeling writer the practical shift is small to describe and large to execute. Lead with what the prescriber should do, use the named phrasing rather than the softer synonym, and write in the active voice, putting the mechanism second and the clinical effect third. Do not restate standard clinical practice and call it a recommendation, and keep in vitro data out of a section that exists to describe clinical significance.
None of that is a matter of taste, which is the part worth registering. A label written the other way around is not merely less elegant. It is, on this guidance's reasoning, less usable at the point of prescribing, and that is a review issue rather than an editorial one. The same logic runs through the label-writing consequences of a subgroup result and through every case where the words in the document, rather than the data behind them, are what a reader acts on.
The draft guidance is at fda.gov/media/182893/download and the QTc guidance at fda.gov/media/170814/download. Both webinar sessions run on 23 September.