01Why a "pyramid" matters

Treating five regulations as a flat "checklist" doesn't work in the field. Three reasons:

Ignore these and the "serious finding" and the "light finding" blur together — making dialogue with the field (MRs, developers) impossible. Hence the pyramid: a structure where higher = larger breach risk.

Important: the pyramid does NOT mean "upper layers must be upheld, lower ones can be ignored." All layers must be upheld as the base assumption. The layering provides priority when multiple findings appear simultaneously, and a basis for explaining how large the consequence becomes if a finding goes uncorrected.

02The 5-tier pyramid

Under the assumption that all must be upheld, higher = greater risk on breach.

Level 1 — Apex

Pharmaceutical Act

Level 2

Advertising Appropriateness Standards

Level 3

Sales-Information-Provision Guideline

Level 4

JPMA Code

Level 5 — Base

Internal SOPs

The "risk on breach" by layer, in table form:

LevelRegulationRisk on breach
Level 1Pharmaceutical ActCriminal penalty / administrative action / approval revocation
Level 2Appropriateness StandardsAdministrative guidance / correction order / disclosure
Level 3HanteiGBusiness-suspension order / mandatory report to management
Level 4JPMA CodeIndustry expulsion / breach disclosure / trust collapse
Level 5Internal SOPsSanctions per each company's own standards

03Three thinking processes for judgment

Holding the pyramid in your head, how do you judge the one case in front of you? Run three thinking processes in order.

Process A — Top-down scan (5 seconds)

On first read, in the first 5 seconds, scan for obvious Level 1 (Pharmaceutical Act) issues:

If a red signal fires, set other discussions aside. Burn down the "layer with the largest consequence on breach" first ── that is the leverage of pyramid thinking.

Process B — Severity within the same layer

Once Level 1 is cleared, look at Level 2 (Advertising Appropriateness Standards). Even within the same layer, separate "clear prohibitions" from "expressions that leave interpretive room."

Process C — Use lower layers as the "alternative" drawer

When you issue NO at a higher level, use lower levels (internal SOPs) as the basis for "this far is OK". This implements the "always offer an alternative after NO" approach from Vol. 1.

Example: "A claim of '100% efficacy' is NG under the Pharmaceutical Act (clear prohibition at Level 2). The internal SOP states 'quoting values from the package insert is OK' (alternative grounded at Level 5). So rewrite it to cite the efficacy rate from the package insert." This is the essence of review — "stop at the top, open at the bottom."

04Pattern classification — a logic tree

Findings that look infinitely varied fall into 4 patterns. As a logic tree:

Is there a problem with this material?
Exceeds approved /
package-insert scope?
↓ YES
Pattern A

Scope exceedance

Efficacy not in PI, off-label use, beyond approved patient range. Concentrated in Level 1-2.

Showing more
than reality?
↓ YES
Pattern B

Exaggerated expression

"Strongest", "only", "definitely" guarantees, excessive headlines, charts emphasizing improvement. Core of Level 2.

Only convenient
information shown?
↓ YES
Pattern C

Missing balance

Efficacy emphasized, AEs understated, selective citation, only good subgroups shown. Core of Level 3.

Inappropriate
scene or medium?
↓ YES
Pattern D

Context problem

HCP content to consumers, what shouldn't be said at a lecture, what shouldn't spread on SNS. Spans Level 3-5.

Most real cases combine multiple patterns. When pointing out: "This is Pattern B (exaggeration) + Pattern C (balance) overlapping". The correction direction becomes much clearer.

05When the pyramid wobbles — edge cases

The pyramid is tidy, but in reality there are moments when the pyramid wobbles. This is where newcomers struggle most.

When in doubt, judge at the higher level: pulling a lower level to declare "safe" in doubt is a weak way of taking risk. Anchoring to "is the higher level upheld?" means you can explain yourself if something escalates later.

06What a veteran's head actually does

Combining everything, the thinking of a veteran reviewer for one material proceeds, in parallel, on this timeline:

5 sec

Process ATop-down scan

Check for obvious Level 1 (Pharmaceutical Act) red signals at top speed. Pre-approval advertising, efficacy out of scope, off-label recommendation ── if red here, others can wait.

30 sec

Logic treePattern classification (A–D)

Apply the 4 patterns ── scope exceedance / exaggeration / missing balance / context. Most cases are combinations, so apply multiple tags.

1 min

Layer IDIdentify which pyramid layer each finding sits on

Is this a Level 1 issue or a Level 4 issue? Prepare the basis to explain "what happens if breached" to the field.

5 min

Process COffer alternatives

From lower layers (mainly internal SOPs and past cases), present "this far is OK". Don't end on NO; hand over a next step toward YES.

Record

TaggingDocument verdict and basis

Tag along two axes: Level (1-5) × Pattern (A-D). Three years from now, this becomes self-study material for a newcomer.

That is what "running five overlapping nets in the head at once" from Vol. 1 contains. The nets are not checked sequentially but run in parallel inside the pyramid. A veteran judges one case in 5 minutes; a newcomer takes 30. The difference is "structured, or memorized as a list".

07Three habits for tomorrow

To land this in daily practice, three habits:

In closing

Whether you memorize five regulations as equal weight or structure them as a pyramid layered by breach risk — the amount of knowledge looks the same. But judgment speed, explanatory power, educational effect, ease of succession — all change by orders of magnitude.

The frame here is simple and usable from tomorrow. Whether you actually reflect it into daily judgments is what divides the skilled three years from now. Newcomer or veteran — try drawing your own "regulatory map in your head" on paper, once.

Next time: how to judge "efficacy claims within the approved scope" inside this pyramid, with concrete examples.