Starting a graph's vertical axis partway up the scale, removing the comparator arm's data, displaying a convenient secondary endpoint prominently while burying the primary one — at first glance these look like technical drafting choices, but they systematically distort prescribing decisions and create risk that patients receive suboptimal treatment. Japan's Ministry of Health, Labour and Welfare (MHLW) promotional monitoring reports have documented this category of violation every year from FY2019 through FY2025, recording more than 24 cases over seven years.
So What / So Why — The Core Problem
The "So what" in a single sentence: the way data is selected, processed, and displayed has itself become a tool of information manipulation. Removing the comparator arm from an efficacy graph, omitting serious adverse events from a safety table, spending two full pages detailing a secondary endpoint while the primary endpoint — which showed no significant difference — gets a few lines of text: each act looks ambiguous in isolation, but as a pattern it constitutes deliberate information manipulation.
The "So why" — why is this fatal — is that the data physicians use as the basis for their clinical judgments becomes contaminated. Doctors decide what to prescribe starting from what is written in papers and what MRs show them on slides. If those slides use a compressed vertical axis or omit the comparator arm entirely, even a highly skilled physician cannot make an accurate assessment. When a drug's efficacy is overestimated, more appropriate alternatives are displaced. When adverse event data is stripped away, monitoring becomes lax and harm goes undetected.
This category is particularly hard to catch because no outright lie has been told. Every data point on the page comes from a real paper or regulatory review report. The problem is what was shown and what was not shown. Unless the recipient independently checks the regulatory review report or the original paper, they have virtually no way of detecting the manipulation.
What / Where / Why / How — What Rule Was Broken and How
What (the four types of violation)
Four main patterns emerge. ① Cherry-picking: selecting only favorable data from multiple sources — deleting unfavorable arms, time windows, or subgroups, and describing only advantageous secondary endpoints in detail. ② Axis, aspect ratio, and reference line manipulation: compressing or stretching the vertical axis, or widening a graph horizontally, to exaggerate or minimize apparent differences. ③ Primary-to-secondary endpoint substitution: presenting a primary endpoint that showed no significant difference in a few lines of text while devoting detailed graphs to a favorable secondary endpoint. ④ Deletion of adverse event and unfavorable information: removing serious items from adverse event tables, omitting pre-medication details, or quoting only part of a paper to make the safety profile look cleaner than it is.
Where (which channels and settings)
Product brochures, presentation slides used by MRs at in-hospital seminars and new-drug briefings, symposium articles and videos on healthcare professional websites, company-hosted webinars, bulk email campaigns, and MR verbal explanations — violations appear at virtually every touchpoint. The pattern is not limited to print; it has been confirmed in video, email, and verbal formats as well.
Why (the psychological drivers)
The monitoring reports identify four main drivers. ① Motivated reasoning: fixing the conclusion — "this drug works" — in advance, then searching only for data that supports it. ② Local rationalization: minimizing each individual item as "just one slide" or "I'll fill in the rest verbally." ③ The sin of omission: "I wasn't asked, so I didn't say" — passive non-disclosure. ④ Responsibility externalization: attributing data selection to outside forces such as "physicians want Japanese-patient data" or "the professor requested this framing." The FY2019 case ②-13 is a textbook example — the responsible employee explicitly explained the subgroup selection by saying "physicians seek Japanese-patient data," a rationale that reveals the structural driver directly.
How (the specific rules violated)
The MHLW Guidelines on Promotional Activities for Prescription Drugs (September 2019) establish as a fundamental principle that "information shall be provided accurately on the basis of scientific and objective evidence" (Section 2, Article 2(2)①), and explicitly state that "when citing figures or tables from original papers, data must not be excerpted, modified, or combined" and that "the scale of graph axes must not be altered, and arrows or reference lines must not be added" (the JPMA Creation Guidelines for Product Information Materials carry the same requirement). The guidelines also require that "whether results used in promotional materials are from primary endpoints, secondary endpoints, or subgroup analyses must be clearly stated." Deleting adverse event information or intentionally withholding unfavorable data violates the guideline's instruction to "take care not to cause misunderstanding by providing only specific information." Article 66 of the Pharmaceuticals and Medical Devices Act (prohibition of exaggerated advertising), the JPMA Code of Practice, and related regulations impose the same obligations.
Case Records — Every Instance
FY2019 (Heisei 31)
This year recorded 14 cases in Category ②, the most of any year across all seven. The range of methods is also striking.
First-line treatment regimens differed substantially between the overall population and the Japanese subgroup (Agent A: 50% overall vs. 20% Japan; Agent B: 30% overall vs. 70% Japan), yet this difference was not disclosed and response-rate graphs were used to claim "higher efficacy in Japanese patients."
Violation: Arbitrary presentation of subgroup results while concealing differences in baseline treatment conditions.
Quoted from report: "The difference may be attributable to the fact that first-line treatment content differed between the overall population and Japanese patients."
A domestic trial had confirmed QT prolongation at a lower dose, yet only high-dose data from an overseas trial was cited to emphasize that "QT prolongation is not a concern." Results from the domestic trial were omitted entirely.
Violation: Safety was emphasized using overseas trial results alone, without disclosing the domestic trial results that were unfavorable to the drug.
Quoted from report: "Because QT prolongation was not observed at 12 mg/day in the overseas trial, such a concern does not exist."
The trial had three arms (placebo, low dose, high dose). The high-dose arm — which showed that increasing the dose provided little additional benefit — was removed, and only the low-dose and placebo arms were graphed.
Violation: Only one or two arms of a three-arm trial were extracted to construct the graph.
The appropriate use guideline explicitly stated that "superiority of this drug over the control arm has not been confirmed," yet Kaplan-Meier curves were used to highlight only the period during which the OS rate appeared higher, asserting superiority.
Violation: Cherry-picking a convenient time window only.
Quoted from report: "Despite the appropriate use guideline stating that superiority of this drug over the control arm has not been confirmed..."
When translating and citing a comparative safety table from the original paper, the "malignant neoplasm" row was omitted — the one adverse event that occurred with this drug but not with the comparator, making it the single most unfavorable safety finding.
Violation: Only the item most seriously unfavorable to the drug was not shown.
The primary endpoint in the original trial was HbA1c "change from baseline," but the brochure substituted "absolute value" and built the graph on that instead. Presenting absolute values does not constitute a valid citation of the original paper or regulatory review report.
Violation: The graph was created using absolute values rather than the primary endpoint of change from baseline (redefinition of the axis metric).
The legend in the video graph was displayed in reverse order. Additionally, the graph's aspect ratio was stretched horizontally, making blood glucose fluctuations appear smaller than they actually were and conveying an impression of "good glycemic control."
Violation: Inaccurate citation and aspect ratio adjustment when reproducing the graph.
Quoted from report: "The aspect ratio had been changed to make the graph wider, resulting in a graph that gave the impression of smaller fluctuations and good glycemic control."
The original paper showed that neither this drug nor the comparator raised blood pressure, but the brochure omitted the comparator graph and presented only this drug's graph, leading to the false impression that only this drug carried no blood pressure risk.
Violation: By omitting the comparator graph, the brochure made it appear as though only this drug had no blood pressure risk.
The primary endpoint (illness duration) was presented in a few lines of text with no comparator data. Secondary endpoints — "change in viral titer" and "time to viral shedding cessation" — were given two full pages of comparative graphs emphasizing superiority. The regulatory review report showed no significant difference on the primary endpoint.
Violation: Only secondary endpoint results that could show superiority were described in detail.
Quoted from report: "There was no significant difference versus the comparator on the primary endpoint, and the way data was presented showed clear selective intent."
From the Phase III adverse event table, several items that the regulatory review report identified as "frequently observed adverse events" — including "ketoacidosis-related events" — were removed.
Violation: Frequently observed adverse events from the clinical trial were not shown in the adverse event table.
Quoted from report: "Multiple items including 'ketoacidosis-related events' had been deleted."
The adverse event incidence rate appeared strikingly lower than competitor products, prompting a review of the regulatory report. This revealed that pre-medication details had been removed from the brochure, and the stated trial objective had been changed from "comparative evaluation of safety" to "verification of pharmacokinetic equivalence" only.
Violation: Pre-medication information necessary for evaluating safety was not disclosed.
The original paper listed serious adverse events by individual name with detailed incidence information, but the brochure grouped multiple items together and omitted noteworthy entries. Numerical inconsistencies were also present. A patient satisfaction survey omitted the note that patients with unknown responses had been excluded from the calculation.
Violation: Inaccurate citation and inappropriate deletion or consolidation of information from the original paper.
Quoted from report: "The original paper lists serious adverse events by individual name with detailed incidence information, but the brochure grouped multiple items together..."
For the primary endpoint of the Phase III international trial (annual asthma exacerbation rate), the overall analysis (approximately 250 patients per arm) was not presented; only the Japanese subgroup (approximately 15 patients per arm) was shown. The responsible employee explained that "physicians seek Japanese-patient data," but the Japanese subgroup happened to show better results than the overall analysis. The brochure contained the same presentation.
Violation: For the primary endpoint, only the results of a subgroup analysis with a small sample size were presented.
Quoted from report: "The response received was that 'physicians seek Japanese-patient data.'"
A post-hoc analysis — "the proportion of patients who had no bleeding events among those who were able to continue weekly dosing" — was presented despite not appearing in the regulatory review report or comprehensive product information. Patients who dropped out of weekly dosing were excluded, introducing a survivorship bias. The slides were never distributed.
Violation: An arbitrary analysis that could cause false superiority impressions was presented.
Quoted from report: "Cases in which weekly dosing could not be continued for any reason had dropped out, making the latter dataset of limited value and liable to create false impressions of superiority."
FY2020 (Reiwa 2)
Results from multiple clinical trials of this drug — trials with different evaluation objectives, eligibility criteria, and doses — were pooled. Primary and secondary endpoint results were presented on equal footing, control arms from non-inferiority trials were omitted, and individual trial sample sizes were not shown. On safety, the Risk Management Plan was not explained and safety data for patients with hepatic impairment was insufficient.
Violation: Data from multiple trials was pooled and presented in a way that gave the impression this drug was superior to other agents.
Quoted from report: "Primary and secondary endpoint results were listed on the same level... overall the materials gave the impression that this drug was superior."
The brochure showed adverse event incidence by treatment duration using a table that included only adverse events with incidence "≥1.0%." The original paper reported events at "≥0.5%," and raising the cutoff caused three adverse events documented in the original paper to disappear from the table.
Violation: By showing only adverse events above a certain threshold, the brochure implied that no other adverse events existed.
Quoted from report: "By limiting adverse event rates to a certain proportion, three adverse events documented in the original paper were not shown in the brochure."
FY2021 (Reiwa 3)
The regulatory review report described a trial in which patients switched from another drug to this drug, but the interview form and comprehensive product information had removed all mention of the prior drug from the study design diagram, creating a figure that implied this drug had been used as monotherapy from the start.
Violation: Despite the trial involving a prior drug, the figure was modified to suggest no prior drug was used.
Quoted from report: "There was no mention of administration of the other drug, creating a figure and table that could lead to the misunderstanding that these were results from switching to this drug alone."
FY2022 (Reiwa 4)
The trial design did not involve a direct head-to-head comparison between this drug and Agent A, yet a "between-group difference estimate" not present in the original paper was added to a shared graph. The same presentation appeared in the comprehensive product information and on the company website, systematically conveying the impression that direct comparison had shown this drug to be superior.
Violation: Data not in the source was added, creating the false impression of a direct comparison.
Quoted from report: "A between-group difference estimate not present in the original paper had been added."
One slide displayed graphs for both primary and secondary endpoints, but the secondary endpoint graph was shown noticeably larger than the primary endpoint graph. During the explanation, efficacy was discussed without mentioning that the larger graph was a secondary endpoint.
Violation: Where the primary endpoint should have been the focus, efficacy was explained using a visually dominant secondary endpoint graph.
Quoted from report: "Efficacy was explained using a secondary endpoint without mentioning that the result was from a secondary endpoint."
FY2023 (Reiwa 5)
In heart failure, both this drug and the comparator are approved only at 10 mg, yet dosing terminology from the diabetes indication was imported: this drug's 10 mg dose was labelled "standard dose" while the comparator's 10 mg dose was labelled "high dose" in text and figures. The cost-comparison slide intuitively framed this drug as standard and the comparator as high-dose.
Violation: Figures that could cause factual misunderstanding were used to present the company's own product favorably.
Quoted from report: "The drug price was explained using the terms 'standard dose' for this drug and 'high dose' for the comparator..."
A patient-facing document contained a diagram suggesting that this systemic transdermal patch provided stronger and broader effects than oral drugs or topical patches. The diagram depicted this drug in a deep color reaching to the extremities with the word "systemic" prominently enlarged, while oral drugs were shown in a pale color covering a narrower area — a visual exaggeration of efficacy.
Violation: Exaggerated expression conveyed the impression of stronger and broader effect (graph manipulation in patient-directed materials).
Quoted from report: "A figure was included that gave the impression of strong and wide-ranging effects."
FY2024 (Reiwa 6)
The source document contained patient and physician survey graphs covering five items related to dosing frequency, administrator, and other factors, but the seminar materials cited only the two items where this drug appeared more favorable than the conventional drug. The source population was severe asthma patients (outside this drug's approved indication) and the survey was conducted overseas, where differences in healthcare coverage systems also influenced responses.
Violation: Excerpting, modifying, or combining data when citing figures from original papers constitutes a guideline violation.
Quoted from report: "Only the items considered to be favorable to this drug compared with conventional drugs were extracted."
In a drug-drug interaction comparison graph distributed by MRs via email, the "no-interaction group" that had been confirmed in the original paper was deleted. The deletion gave the impression that this drug had fewer interactions.
Violation: Constitutes data excerption, modification, or combination when citing figures.
Quoted from report: "This portion had been deleted from the graphs in the series of materials sent by the company representative."
FY2025 (Reiwa 7)
For an international Phase III trial, an MR explained verbally — and the brochure stated — that "this drug is the only treatment for [disease] that has demonstrated improvement in long-term prognosis." However, the data supporting improved long-term prognosis came from "time to all-cause death," a secondary endpoint — not the primary endpoint.
Violation: Results from a secondary endpoint were used to make the exaggerated claim of being "the only" treatment.
Quoted from report: "Claiming that this is the only drug to have demonstrated improved long-term prognosis constitutes an exaggerated expression."
Learning from Past Incidents ── Map of 9 chapters
- Part 1: Data Misuse and Manipulation That Risk Factual Misrepresentation
- Part 2 (this chapter): Selective Data Extraction, Manipulation, and Presentation
- Part 3: Claims Without Evidence / Unreliable Data
- Part 4: Presenting Unapproved Indications and Dosage Regimens
- Part 5: Exaggerated Claims
- Part 6: Efficacy-Only Promotion / Safety Information Neglect
- Part 7: Disparagement of Competing Products
- Part 8: COI Non-Disclosure
- Part 9: Other Improper Sales Practices
- The way data is "selected" constitutes information manipulation. Removing the comparator arm from an efficacy graph, deleting serious adverse events from a safety table, and reserving detailed graphs exclusively for secondary endpoints are all technically truthful acts — yet they systematically distort the prescribing decisions of the recipient. Unless physicians independently check the regulatory review report, detection is nearly impossible.
- Axis manipulation and aspect ratio changes are invisible to the eye. Compressing the vertical axis, widening a graph horizontally, or reversing a legend all look like "layout choices" at first glance, yet they convey a materially different impression of effect size or variability than the original paper. The FY2019 report alone recorded two such cases (②-6, ②-7).
- Secondary endpoint substitution recurs as a structural pattern. Even when a drug's primary endpoint shows no significant difference, secondary endpoints or Japanese subgroups sometimes yield favorable results. Describing only those results in detail has been confirmed in every year from FY2019 through FY2025, pointing to a design problem in material authoring and review systems.
- Ministry of Health, Labour and Welfare, "Guidelines on Promotional Activities for Prescription Drugs" (September 25, 2019)
- Ministry of Health, Labour and Welfare, "Promotional Activity Monitoring Project for Pharmaceutical Products" — Annual Reports FY2019 through FY2025
- Japan Pharmaceutical Manufacturers Association (JPMA), "Creation Guidelines for Product Information Materials for Prescription Drugs"
- Japan Pharmaceutical Manufacturers Association (JPMA), "Code of Practice"
- Pharmaceuticals and Medical Devices Act, Article 66 (Prohibition of Exaggerated Advertising) and Article 68 (Prohibition of Advertising Unapproved Drugs)