A pre-MLR content check is the gate that runs before your medical, legal, and regulatory reviewers ever open the file. It does not approve anything. It sorts: this asset is clean enough to move, this one needs a human eye on three specific lines, this one cannot proceed as written. In practice that resolves to three verdicts, Cleared, Review, or Blocked, and the entire value of the exercise depends on what those verdicts are measured against. The honest answer is that they are measured against the label of record, the SmPC or Prescribing Information, and not against a brand book. That distinction is the difference between a check that holds up under enforcement and one that is decorative.
This piece walks through what the verdict actually returns, why label grounding beats style grounding, what the rule engine looks for, and where the line sits between flagging and adjudicating. A pre-MLR check should make the human review faster and more focused. It should never pretend to be the human review.
The three verdicts, and what they are not
Cleared, Review, and Blocked are triage states, not approvals. Cleared means the automated checks found nothing that trips a rule: claims trace to the label, required safety information is present and balanced, and the language stays inside what the indication supports. Review means something needs a person, a claim that is close to the label but not exact, a fair balance ratio that looks thin, a superlative that may or may not be substantiated. Blocked means the asset contains something that should not advance in its current form, an off-label benefit claim, a missing contraindication, an indication the product does not hold.
The reason this matters is that MLR review has become the choke point in pharma content production, not because reviewers are slow but because volume has outrun the process. Industry commentary describes manual MLR review taking weeks per asset and notes that teams often rely on a small number of expert reviewers, which concentrates risk and creates burnout (Viseven). A pre-check earns its place only if it removes noise before that scarce human attention is spent. Cleared assets move with a clean trail. Review assets arrive with the exact lines flagged. Blocked assets never waste a reviewer's afternoon on a problem a rule could have caught in seconds.
What the three verdicts are not: they are not a regulatory sign-off, they are not a substitute for the reviewer's judgment, and they are not a guarantee. They are a defensible, repeatable first pass that tells your team where to look.
Why label grounding beats a brand book
A brand book tells you how the product should sound. The label tells you what the product is allowed to say. These are not the same document, and only one of them is what a regulator will hold you to.
In the United States, every prescription drug advertisement must present a true statement of information relating to side effects, contraindications, and effectiveness, and must not be false or misleading on those points (21 CFR 202.1). FDA's enforcement posture makes the grounding explicit: promotional materials must be consistent with the FDA-approved prescribing information, and the agency cites misalignment between the benefits an ad highlights and the product's approved indication as a violation. In its September 2025 wave of more than 100 enforcement letters, FDA targeted overstatements of efficacy and failure to adequately present risk, faulting ads where benefit claims were not supported by the approved labeling (Latham & Watkins).
Europe draws the same line even more bluntly. Under Article 87 of Directive 2001/83/EC, all parts of the advertising of a medicinal product must comply with the particulars listed in the Summary of Product Characteristics, and the advertising must encourage rational use, present the product objectively without exaggerating its properties, and not be misleading (EUPATI). The SmPC is described as the basis of information for healthcare professionals on how to use a medicine safely and effectively, and as the document from which the package leaflet and other materials are prepared (EUPATI).
So a pre-MLR check that scores against a brand book is checking the wrong thing. It might tell you the headline is on-tone and the logo is in the right corner. It cannot tell you whether the headline implies a benefit the label does not grant. Label grounding inverts the priority: the SmPC or PI is the source of truth, every claim is matched back to it, and tone is a secondary concern the human reviewer can still own.
Inside the rule engine: fair balance, ISI, claims, language
A pre-check runs several passes, each anchored to the label and to the regulations that govern promotion.
Fair balance and the safety information pass. The check looks for whether risk is present and whether it is presented with weight comparable to benefit. The regulatory standard is specific: an ad is not considered to lack fair balance when the presentation of true information on side effects and contraindications is comparable in depth and detail to the claims for effectiveness or safety (21 CFR 202.1). So the engine checks for the presence of Important Safety Information, that contraindications and major warnings carried in the label appear, and that the benefit copy does not dwarf the risk copy in scope or prominence. Missing ISI or a one-sided benefit-to-risk ratio is exactly the kind of finding FDA's recent letters cited.
Claim matching. Every substantive claim in the asset is matched against the label and against your approved claims library. A claim that maps cleanly to an indication or to substantiated label language passes. A claim that is adjacent, broader, or comparative without support gets flagged. A claim for a use the product does not hold is the bright line: promoting a drug inconsistent with its approved labeling is the definition of off-label, and it is what misbranding enforcement turns on.
Language pass. This is where superlatives, implied superiority, absolute safety language, and tone that overstates efficacy get caught. The EU standard against exaggeration and the US standard against misleading presentation both live here. The language pass is also where the check distinguishes a stylistic preference, which belongs to the brand book and the human reviewer, from a regulatory exposure, which belongs to the rule engine.
The output of all of this is not a grade. It is a set of located flags, this line, this rule, this label reference, rolled up into one of the three verdicts.
A worked example: Varigel
Take a fictional product, Varigel, indicated for moderate plaque psoriasis in adults. A brand team submits a digital HCP banner for pre-check. The copy reads: "Varigel delivers fast, lasting clearance, the new standard in psoriasis care." There is a small-print line at the bottom referencing safety.
The pre-check runs its passes against the Varigel label of record. The claim matching pass flags "lasting clearance," because the label supports a measured response endpoint at a defined timepoint, not an unqualified durability claim. The language pass flags "the new standard in psoriasis care" as an unsupported superiority claim and "fast" as an efficacy characterization not framed by the label's onset data. The fair balance pass flags that benefit copy spans the headline and subhead while ISI is reduced to a single small line, a depth-and-detail imbalance of the kind the regulation addresses. Crucially, the indication itself, moderate plaque psoriasis in adults, matches the label, so this is not off-label use, it is overreach within the right indication.
Verdict: Blocked. Not because a machine decided the campaign is non-compliant, but because the asset as written carries findings that should not advance: an unsupported superiority claim and a fair balance shortfall. The reviewer who opens this file does not start from a blank page. They start from three located flags, each tied to a label reference, and they make the call. Maybe "lasting clearance" becomes "clearance maintained through week 16" with the citation. Maybe "the new standard" comes out entirely. The human adjudicates. The check just made sure the human was looking at the right three lines in seconds instead of reading the whole asset cold.
If the same banner had matched the label on every claim and carried balanced ISI, the verdict would have been Cleared, and it would have reached MLR with a clean, auditable trail showing why.
Flagging, never adjudicating: the line that keeps you safe
The most important design decision in a pre-MLR check is the one that says the tool does not decide. It flags, it explains, it grounds each flag in a label reference and a rule, and then it hands the asset to a person. Industry guidance is consistent on this: AI in MLR is an addition, not a replacement, and human oversight remains necessary across compliance and review (Viseven).
There are good reasons beyond philosophy. Promotional review involves judgment about context, audience, and substantiation that a rule engine should inform but not own. A false Cleared from a tool that adjudicated would be a liability with your name on it. A Blocked verdict that a human can overturn with documented rationale is a safe, auditable workflow. The verdict is a recommendation with evidence attached, and the evidence is what makes the human review faster rather than redundant. This is also why a pre-check that supports 21 CFR Part 11-style audit trails matters: the value is not just the verdict, it is the record of what was checked, against which label, and who made the final call.
Where this leaves you
A pre-MLR content check is worth running only if it is grounded in the right source and honest about its role. Grounded in the label of record, the SmPC or PI, it catches the claims, fair balance gaps, and off-label exposures that regulators actually act on. Honest about its role, it flags and explains, then steps back so a qualified reviewer adjudicates. Cleared, Review, Blocked is not a verdict on your campaign. It is a verdict on where your reviewers should spend their attention.
That is exactly what Juncture Pre-Check does: a label-grounded rule engine that runs fair balance, ISI, claim matching, and a language pass against the SmPC or PI of record, returns a Cleared, Review, or Blocked verdict with each flag tied to a label reference, and hands every asset to your MLR team with the trail intact. It is 21 CFR Part 11-supporting, and it never adjudicates. It just makes sure the humans who do are looking at the right lines first.
Sources
- 21 CFR 202.1, Prescription-drug advertisements (brief summary, fair balance, false or misleading), Legal Information Institute
- FDA Begins Crackdown on Direct-to-Consumer Pharmaceutical Advertising, Latham & Watkins
- Directive 2001/83/EC, Article 87 (advertising must comply with the SmPC), EUPATI Open Classroom
- Summary of Product Characteristics (SmPC), basis of information for healthcare professionals, EUPATI Open Classroom
- MLR Review Process in Pharma: Challenges and Solutions, Viseven