Making the bird's-eye view a habit — reading past failures as a mirror

When you review a single piece of promotional material, your attention narrows to that case. That narrowing is necessary. But when you step back and read across multiple years of deviation records, something becomes visible that was not before. "The secondary endpoint had better results." "Physicians want Japanese data." The same words appear, year after year, drug after drug, reviewer after reviewer. The bird's-eye view is the habit of placing yourself outside any individual review and asking whether today's judgment is traveling a familiar route. This final installment asks what it takes to make that habit stick.

So what / So why — the core of this perspective

The Ministry of Health, Labour and Welfare's monitoring program for pharmaceutical sales information activities (including the advertising activity monitoring program) has run continuously since 2019. The fiscal year 2025 report noted that "this is the ninth year since the program began." That same report stated: "Inappropriate sales information provision activities continue to be observed, including cases using exaggerated expressions such as 'the only drug that improves ~' based on secondary endpoint results, cases that described efficacy while neglecting to provide safety information, and cases where only the portions favorable to the company's own product were excerpted when comparing with competitors' products." Nine years in, the same categories persist. That fact is what this perspective is built on.

Case-by-case review tends to stop at questions like "is this data accurate" and "is there a basis for this claim." Yet deviations have kept slipping past exactly those questions. The problem lies one step before accuracy — in choices about what to show, what to foreground, which endpoint to enlarge. Those choices are easy to miss when you are looking at a single case. Read across multiple years of records and a specific structural gravitational pull comes into focus.

The bird's-eye view means making that cross-reading a habit, then asking whether the case in front of you today carries "the same shape as the recorded pull." Series opener vol. 01 "Standing as a custodian of evidence" posed the same question — "Are you a persuader or a custodian of evidence?" — and the bird's-eye view brings that question into every review, using past failures as the mirror. That is what habituating it looks like.

Inside the perspective — What / Where / Why / How

What (what to look at) — Extract the shape of each gravitational pull from past deviation records by category. The pull toward foregrounding secondary endpoints while burying primary ones. The pull toward showing only a favorable subgroup rather than the full population. The pull toward generating a "the only drug that..." modifier from secondary endpoint results. Name each one and keep it close. From there, a trained eye develops that asks "which pull does this material resemble?" during any given review.

Where (where to apply it) — Add a single "bird's-eye question" line to the end of your normal review workflow. At the bottom of any checklist, write: "Which recorded pattern does this judgment most resemble?" That is enough. Setting aside time once a month to re-read one past incident from your organization's collection builds the same effect over time.

Why (why it matters) — Individual case review stops at the accuracy question. But the gravitational pull operates outside that question. Without cross-reading, you can be drawn in by the same pull repeatedly. What nine years of records demonstrate is that there are structural repetitions that individual honesty alone cannot prevent.

How (how to make it stick) — Rely on structure, not willpower. Add one line to the review checklist. Schedule a monthly incident re-read. Hold a short incident read-aloud with the team. None of these require elaborate systems. Embedding the question in a process means the bird's-eye view activates even on low-energy days.

Stepping into someone else's position

Consider the position of a prescribing physician.

A physician has received multiple product presentations from the same company about the same drug over six months. The numbers in every presentation were accurate. Each time, a graph showing favorable secondary endpoint results filled most of the slide. Efficacy dominated ninety percent of the conversation. Safety was covered in a single slide listing contraindications. Questions about the primary endpoint were answered when asked, but the primary endpoint data was never shown as a chart. The physician added the drug to their prescribing practice with an impression of it as something that works well.

After a patient experienced an adverse event, the physician looked at the regulatory review report for the first time. There they found that the primary endpoint had not reached statistical significance, that only the secondary endpoint analysis had produced a positive result, and that several important risks were listed under safety. The feeling was not "this contradicts what I was told" but "what I was told was incomplete."

The physician bears no responsibility here. The time and attention available for any prescribing decision are finite, and checking the regulatory review report for every drug is not realistic. The design of the information package was on the side that created and delivered the materials. Only the maker can control what image remains in the recipient's mind. The capacity to imagine what that physician's prescribing decision was based on is the starting point for the bird's-eye view. Past incidents are also, in a real sense, records of those physicians.

Reading past deviations as a mirror

Place two incidents side by side.

2019 (Ministry of Health, Labour and Welfare Advertising Activity Monitoring Program Report, Project No. 000509783) — At an in-hospital product presentation for a bronchial asthma treatment, results from the full analysis population for the primary endpoint (annual asthma exacerbation rate) in a phase III international joint trial — roughly 250 patients per arm — were not presented. Instead, only a subgroup analysis of Japanese patients (approximately 15 per arm) was included in slides and the pamphlet. When asked why only Japanese data was shown for the primary endpoint, the response was: "Physicians want Japanese data." The case was recorded as appearing arbitrary because the subgroup produced better results than the full population, despite the smaller sample size. (See Anatomy of a Failure vol. 02 "Selective extraction, manipulation, and presentation of data and graphs")

2025 (Ministry of Health, Labour and Welfare Sales Information Provision Activity Survey Report, Project No. 001520054) — For a diabetes drug, company representatives explained in online group interviews and a product pamphlet — in the context of a phase III international joint trial — that "this drug is the only treatment for ** disease that has demonstrated improved survival prognosis." However, the survival prognosis improvement data came from a secondary endpoint (time to all-cause mortality), not the primary endpoint. The claim of being "the only" drug to improve survival prognosis, based on a secondary endpoint result, was judged to constitute an exaggerated expression. (See Anatomy of a Failure vol. 05 "Exaggerated and definitive expressions")

Viewed from above, the same structure surfaces across a six-year gap. A conclusion came first, and data was selected to support it. In 2019 it was "a subgroup that outperformed the full population." In 2025 it was "a secondary endpoint that outperformed the primary." The object of selection differed; the gravitational pull was the same. If the 2025 material creators had kept the 2019 incident close as a mirror, would the question "is what I am about to select the same shape as that recorded choice?" have engaged? That is precisely what the bird's-eye habit is asking about.