Ch.11 Abstract collections | JPMA Guide Commentary | Advertising Regulation in Japan (Promotional Material Review) and Artificial Intelligence

A literature abstract collection is a single resource that gathers summaries of several original research papers into one bound piece. Instead of handing over one reprint, the maker edits and binds the abstracts of multiple papers. Convenience comes at a price: because an act of editing sits in the middle, the maker's intent slips in more easily. It is no accident that this resource closes Part III. Within the family of materials provided on request — abstract collections, reprints, proceedings — the literature abstract collection is the one where the editor's hand reaches furthest, and so the guardrails around it are the heaviest.

One question runs through this resource: "Is this a summary collection that answers a healthcare professional's request, or a promotional piece wearing the costume of a summary?" Bind the same papers, and depending on how you pick, order, and color them, the two become entirely different objects. The guide imposes detailed conditions precisely to protect that boundary by structure rather than by trust.

01Standing — Why "on request" is the premise

A literature abstract collection is a resource provided in response to a healthcare professional's request. It is not something to hand out proactively. And the format itself must not give the impression that proactive distribution is the premise. The spirit of the foreword bears directly here. Just as the product information summary is a subordinate piece that "supplements" the package insert, the abstract collection is a passive tool that answers when a physician comes to seek information.

Why must it be passive? Because a collection of third-party papers looks, at first glance, like neutral academic information rather than a company's claim. Use that neutral guise to push a product, and you commit the violation the foreword fears most: "true, yet misleading." The papers may be genuine, but if the binding is advertising, the whole is advertising.

02Quality of material — Peer-reviewed originals only

A collection may be built only from abstracts of original papers that have passed rigorous peer review. Reviews, narrative reviews, articles, and contributed pieces are excluded. This drops the discipline of the evidence hierarchy straight into resource design.

Peer review is the watershed that separates the quality of scientific claims. An original paper that has been peer reviewed has had its methods and results pass third-party scrutiny. A review, by contrast, is the author's argued selection from existing studies; the author's interpretation and choices are necessarily baked in. Bind reviews and you get "a bundle of interpretations"; bind originals and you get "a bundle of primary data." What a collection demands is the latter.

The temptation to mix in reviews is easy to understand: their conclusions are crisp and often carry a summary favorable to the product. But that crispness is itself the product of the editor's interpretation, not primary evidence. Putting a review into an abstract collection is equivalent to slipping unverified opinion into a bundle of verifiable fact.

03Selection criteria — Securing objectivity by stating it

Which papers to include must be decided by objective criteria, and those criteria must be stated in the resource — for example, by recording that a neutral editorial committee selects papers against clear, explicit standards.

This is the vital point of the collection. Paper selection is the quietest and most effective stage at which to manipulate impression. Gather only the convenient papers and drop those with unfavorable results, and every individual summary stays accurate while the whole forms a slanted picture. The "ban on deliberate omission" common to Part III's lists and comparisons takes, in the collection, the concrete form of "stating the selection criteria."

Why writing the criteria down acts as a guardrail

Once selection criteria are recorded in the document, those criteria become verifiable from outside. To the question "under this criterion, shouldn't that unfavorable paper also be included?", the maker now has to answer by the criterion. Forcing the standard into writing itself restrains arbitrary cherry-picking. This applies one of the foreword's three pillars — "securing verifiability by structure." By making the procedure disclosed, the fairness of the result can be inspected after the fact.

Same set of papers, different impression. Suppose a field has ten high-quality original papers, of which seven show benefit and three show no effect or harm. Build the collection from the seven favorable ones only, and although each summary is perfectly accurate, the reader takes away the picture "this drug generally works." Stating the objective selection criteria exists to block this trick of "assembling a false whole from accurate parts."

04The ban on processing — Faithfully reflecting the original's intent

Each summary must faithfully reflect the true intent of the original paper. No biased processing. No additions of things the original does not state. No claims in the summary that go beyond the original.

Summarizing is, by nature, an act of cutting. In cutting, a judgment about what to keep and what to discard always enters. If that judgment leans systematically toward what favors the product, the summary stops representing the original and becomes a substitution of the company's claim. The same discipline as Chapter 1's "quote guidelines verbatim, reflect the intent faithfully" is imposed here on original papers.

The condition "no additions beyond the original" weighs especially heavily. Add to the summary an implication or clinical value the original never showed, and that is a fresh claim built on the original as a stepping stone — groundless exaggeration. The summarizer's job is translation, not invention.

05Limits of content — Three lines never to cross

The content of a collection must not cross the following three lines. Each is the collection's form of a prohibition that runs through the whole guide.

It must not lead to recommending off-label use. Even if an original paper covers uses or indications outside the approval, it must not be bound in a context that recommends them. The approved content is the master text, and the collection cannot exceed it by a single character.
It must not lead to disparagement. Do not center the collection on papers unfavorable to competitors, or arrange comparator results so as to denigrate them.
It must not lead to recommending uses that conflict with the precautions and related information. Do not, through summaries, suggest a manner of use that contradicts the warnings, contraindications, or precautions in the package insert.

Color as a hard-to-see manipulation

The coloring of abstracts must follow the original or be more restrained than the original. Take figures the original printed in black and white, and color only the product-favorable data in the collection, and you steer the reader's gaze without altering a single character of text. This shares the thinking of Chapter 1's "no emphasis by color or bold." Emphasis is achieved not only by text but by visual effect. By specifying even the tone of color, the guide blocks misdirection through appearance.

06Cover and bibliography — Declaring passivity and verifiability

The cover must state, in gothic type of at least 10 points, that the package insert should be consulted and that the resource is provided on request. The type size is fixed because these are function, not decoration. The line "on request" is a declaration that this resource is a passive tool; the line "consult the package insert" is the caveat that a summary collection is no substitute for the master text.

The bibliographic details of each paper must be recorded accurately. The selection criteria for papers and the scope of the company's involvement must also be stated, along with the company name. If the bibliography is accurate, readers can go to the originals and judge for themselves whether the summaries are sound. If the scope of involvement is disclosed, readers can read while weighing the conflict of interest. Both are requirements for creating a state that can be verified after the fact.

Do not treat "stating the scope of company involvement" lightly. Who funded it, who selected the papers, who wrote the summaries? Without this information, readers receive it as a neutral academic collection. Concealing involvement behind a neutral guise breaks the integrity of the whole resource, even if every summary is accurate.

Point	Permitted construction	Construction that crosses the line
Material	Abstracts of peer-reviewed originals	Mixing in reviews, articles, contributions
Selection	Objective criteria stated, neutral choice	Favorable papers only, criteria undisclosed
Summary	Faithful to the original's true intent	Biased processing, additions beyond original
Color	As original or more restrained	Favorable data emphasized by color
Provision	On request; passivity shown by format	Format premised on proactive push

In closing

The difficulty of the literature abstract collection lies in the fact that every part is genuine. Peer-reviewed papers, accurate bibliographies, faithful summaries — inspect them one by one and there is no flaw. That is exactly why the guide disciplines the assembly rather than the parts. What you select, how you order, which color you apply, how you hand it over. Bias lives not inside a summary but in the gaps between summaries.

Stating the selection criteria, providing on request, respecting the original's true intent — all of these translate the foreword's duty to "convey accurately without misleading" into the format most easily manipulated: an edited compilation. Assembling a false whole from accurate summaries is not permitted. What prevents it is not the maker's conscience but the structure of disclosed criteria and verifiable bibliography.

Case studies — the "it isn't written, so…" mindset, and how each lands

Abstract books serve as the primary record of presented research. Four schemes — disguising article type, obscuring author involvement, concealing corporate drafting — all rest on the same premise: looking like real science is enough.

GRAY 1 — Straying from the spirit (untouched by the wording, but against the foreword's spirit)

The scheme: "Including review articles isn't banned. So I'll use a single unified format for both reviews and original research, with no label distinguishing them. Readers will read it as a body of primary evidence — we're just being consistent with our layout."

Verdict: Intentionally unifying formats so readers conflate reviews with primary research causes misreading of evidential weight — directly against the foreword's instruction to prevent misunderstanding. Reviews and original papers carry different authority as evidence. Minimum acceptable practice: label each entry with its article type in the table of contents or in a header field.

GRAY 2 — Exploiting the gap (deliberately working a loophole to invite misreading)

The scheme: "Reprinting abstracts from a journal is permitted. So we write the summary in-house, have it listed as an abstract for an external CME symposium, then get it placed in that symposium's journal supplement. Now we can circulate it as a reprint."

Verdict: Routing company-drafted content through an external venue to acquire the appearance of a journal reprint is a textbook gap exploit. The distribution format is reprint; the content's origin is a commercial department. Deliberately blurring the boundary between academic literature and corporate material breaches the JPMA Code's independence principle regardless of the publication vehicle used.

GRAY 3 — Violating higher norms (arguably "not in the text," yet almost certainly out under the PMD Act, Fair Advertising Standards, and the JPMA Code)

The scheme: "We prioritised abstracts with data favourable to our product and omitted unfavourable presentations 'due to space constraints.' No rule specifies the selection criteria, so the discretion is ours."

Verdict: An abstract book that systematically removes presentations unfavourable to the compiler's product loses its identity as a neutral account of scientific discourse. The JPMA Code demands accurate and balanced information; using editorial discretion as cover for outcome-correlated exclusion is a breach of that requirement. "Selection discretion" does not immunise structured suppression of negative data.

BLACK — Breaching the explicit text (a plain violation of the provision itself)

The scheme: "We drafted the review, listed an external physician as the author, paid them a fee, and had them check the content. The author field shows only the physician's name. They did review it, so calling them the author isn't a lie."

Verdict: A company-drafted manuscript published under a physician's name while concealing the corporate origin is ghost authorship. It fails the ICMJE authorship criteria, which require substantial intellectual contribution to conception and drafting — not just a content check. The JPMA Code mandates transparent disclosure of both author involvement and company involvement. Omitting the in-house drafter from the byline is a plain breach of the disclosure obligation; fee payment and a review pass do not confer authorship.

Improper intent by implication — three scopes of "insinuating" without stating it

A literature abstract compendium can satisfy both formal requirements — stated selection criteria and peer-reviewed original articles only — and still produce a false overall picture if the design of the criteria or the ratio of included versus excluded studies is skewed. The grey zone lives here.

Implication 1 — Local: carried in a single word or sentence

How it is built: The stated selection criteria include the sentence: "Studies with statistically significant results were prioritised for inclusion."

Intent that can be read: "Statistically significant" sounds like a scientific filter. In practice, statistically significant results skew toward trials with larger sample sizes or designs favourable to the company's product. Framing this as a criterion makes the exclusion of non-significant — often unfavourable — results appear methodologically legitimate.

Verdict: Statistical significance alone is not a valid selection criterion: underpowered studies may be non-significant for reasons of sample size, and a non-significant result can carry important clinical meaning. Selection criteria must be built on study design quality, relevance to the indication, and comparability — not on whether results favour a particular direction.

Implication 2 — Adjacent: read across neighboring clauses

How it is built: Each abstract ends with a "Study limitations" sub-section. The company product's trial limitation reads "Relatively small sample size." The competitor trial's limitation reads "Inadequate control of confounders; interpretation of efficacy findings requires caution."

Intent that can be read: Both statements are drawn from the original papers. But limiting the company trial's limitation to a quantitative constraint (sample size) while expressing the competitor trial's limitation as a question about the reliability of its efficacy conclusion guides the reader's comparative judgement. The evaluation is created by editorial choice, not by the data.

Verdict: Limitations must be transcribed for all studies at the same level of detail and using the same structure. If the editor selects which limitations to highlight and at what depth, the resulting summaries are accurate individually but dishonest in aggregate. Apply a uniform template for limitations across all included studies.

Implication 3 — Whole-context: arising from the document as a whole

How it is built: Stated selection criteria are "randomised controlled trial, peer-reviewed, minimum 12-week treatment duration, indication matching the approved use." Ten papers on the company product and six on competitors meet all criteria. An editorial board adds "trial scale" as a secondary criterion and includes 9 of 10 company-product papers and 2 of 6 competitor papers, for a final compendium of 11 papers from a pool of 16.

Intent that can be read: The stated criteria are documented; every abstract is accurate. But the secondary criterion was applied in a way that yielded a 90% inclusion rate for company-product studies and 33% for competitors. Readers form the impression that the evidence base is anchored in the company's own trials, without seeing the selection asymmetry.

Verdict: Stating selection criteria and applying them neutrally are separate requirements. When a supplemental criterion, however defensible in isolation, produces a large imbalance in inclusion ratios, the criteria design itself must be reconsidered. Fairness of selection should be verifiable from the inclusion ratios in the final compendium, and those ratios should not diverge markedly between the company's own studies and others.