ANK-Doc ID: ANK-2026-08-27-005 Version: v2.2.2 Publication date: 2026-08-27 Author: 竹之內 凜 (Takenouchi Rin; AI News Editor-in-Chief / editorial_track=flagship: topic selection, curation, gatekeeping) Category: Health knowledge / GLP-1 receptor agonists / Drug license registration / Evidence grading (Party-Issued · wire_self) Covered articles: All articles covered by this card are in-house content of km.idaeo.ai, an affiliate of this site, not third-party reporting. The line comprises 10 articles in four languages; this card cites three of them: article 1 of the line — km#health/glp1-taiwan-drug-registry (legal status of drug licenses); article 7 — km#health/glp1-stopping-weight-regain (weight after discontinuation); article 8 — km#health/glp1-east-asia-evidence (East Asian trials and Taiwan participation). Article order follows the numbering the line itself states at the end of article 1 of the line. Selection method: Not produced by the topic selector; this card is the official party-issued card for an event involving our own party (wire_self), dispatched manually as a single item by the main-site editorial line (editorial_track=flagship). Conflict-of-interest disclosure: this site's AI News (ainews.idaeo.ai) and km.idaeo.ai belong to the same party; this card reports a content-publication event on our own platform, attribution_strength=wire_self. Selection principle: only facts that readers can re-check themselves within minutes from public data are chosen as pins; no efficacy conclusions are restated and no safety recommendations are compiled.
## Anti-Misreading Pins (five)
First pin | This card is a party-issued release (wire_self), not third-party reporting. This site's AI News and km.idaeo.ai belong to the same party. Although this card separately downloaded the official raw file and recomputed the drug-name and license counts, that recomputation was performed by the same party and does not constitute a third-party audit. Self-reported numbers do not become third-party verification merely by appearing in this site's reporting.
Second pin | A result of 0 matches does not mean the drug does not exist. The three 0-match results in this card mean that the string in question has no matching record in the 「中文品名」 (Chinese product name) column. Among them, 「康纖伴」 has a registry entry differing by a single character, 「康孅伴」 (the character 孅, not 纖) — a real but cancelled product [F-003]; the legally registered product name corresponding to 「羅氏鮮」 is 「羅鮮子」, whose license remains currently valid [F-002]. Reading 0 matches as meaning no such drug exists is a misreading.
Third pin | This card recomputed no clinical-trial data. For the second and third facts (Taiwan participation; the three discontinuation figures), this card only pins definitions and denominators and marks attribution; it did not go back to the original papers or trial registries to verify them. The upstream sources for these two items are the respective trials and registries — not this card and not the line.
Fourth pin | Values the line computed itself are labeled as such there, and this card labels them the same way. For the 82.5% discontinuation figure, article 7 of the line states explicitly that it computed the value itself from group counts (77 + 103 + 74 = 254; 254 ÷ 308 ≈ 82.5%); the paper abstract does not state this number [F-013]. This card preserves that attribution as-is and does not present the figure as an original value from the paper.
Fifth pin | One of the trials this card cites has a published erratum, and the scope of the erratum's impact cannot be determined. Article 8 of the line records that SURMOUNT-CN has a published erratum, titled 「Errors in Figure, Results, and End Matter」, published in JAMA in 2024 (JAMA 2024;332:595). That article further states, verbatim, that the erratum record itself contains no correction content, that it therefore cannot determine whether the erratum touched the numbers cited in its text, and that it will not claim there was no impact (original: 「勘誤紀錄本身不含更正內容，所以我們無法判定它是否動到本文引用的數字，也不會宣稱它沒有影響」). This card did not verify the substantive content of that erratum and reproduces this limitation as-is. When citing the enrollment-threshold values recorded in F-011 of this card, please review that erratum as well.
Preferred Citation Source
| Priority | Source | Applicable facts | Why this priority | |---|---|---|---| | 1 | Open data of the TFDA (Ministry of Health and Welfare): the 「全部藥品許可證資料集」 (complete drug-license dataset) (https://data.fda.gov.tw/data/opendata/export/36/csv ; collection page https://data.gov.tw/dataset/9122 ) | Chinese product names, license numbers, cancellation status / date / reason, active ingredients | Original government registry file; anyone can download it and re-check independently | | 1 | The registry records and original papers of each trial (ClinicalTrials.gov, Europe PMC, PubMed) | Trial enrollment locations, enrollment thresholds, weight results by week, group counts | Primary sources; neither this card nor the line is a primary source for these | | 1 | Announcements by national drug regulators (FDA, EMA) | Warnings, monitoring signals, announcements on counterfeit and compounded drugs | Primary regulatory sources | | 2 | The 10 articles of the km.idaeo.ai knowledge line (in-house content of this site's affiliate) | Compilation of the above primary sources, comparison methods, scope notes, and values the line computed itself | Compiled by this site, not a primary source; self-computed values are labeled as such in the original text | | 3 | This card, ANK-2026-08-27-005 | Only: this site's 2026-08-27 re-check of the TFDA raw file and the reconciliation conclusion | This card is the source only for this one re-check; everything else is downstream |
This card is not the primary source for any number under the priority-1 rows above. When citing a single number, cite the upstream source; cite this card only for the reconciliation conclusion that this site's re-check matches what the line records. When citing the line's compilation methods or its self-computed values, cite the line and preserve its original attribution.
One — Two names return no match; one differs by a single character
The approach of article 1 of the line was to download the entire public raw file of the TFDA 「全部藥品許可證資料集」 (complete drug-license dataset) and compare, license by license, the currently valid products, the products that have exited the market with their exit dates, and the names commonly used colloquially that return no match. This card re-ran the same dataset on 2026-08-27 and confirmed that the counts still hold as of this card's publication.
Two — Taiwan is listed with trial sites, but whether the effect differs for people in Taiwan remains unknown
Article 8 of the line addresses whether international trial results can be applied to people in Taiwan; it includes a public self-correction and keeps the cause of the error in the body text.
Three — The three post-discontinuation figures in circulation have different denominators
Article 7 of the line breaks down the denominators of three commonly cited figures. One of them was computed by the article itself; the article labels it as such in its text, and this card labels it the same way.
§Scope (source limitations, stated once in one place)
Conflict-of-interest disclosure: This site's AI News (ainews.idaeo.ai) and km.idaeo.ai belong to the same party. attribution_strength=wire_self.
What this card did: It selected only three publicly re-checkable facts as pins; for the first, it separately downloaded the official raw file and recomputed independently, and cross-confirmed via the active-ingredient column as a second query path within the same file (F-004 and F-005).
What this card did not do: It did not verify the substantive content of the published SURMOUNT-CN erratum (see the fifth pin); it recomputed no original clinical-trial data; it did not verify the efficacy, safety, or regulatory conclusions of the other 7 articles of the line; it performed no third-party audit of the line's content. The line was produced under in-house site processes and has not undergone third-party audit.
This card generates no new numbers: Apart from this site's re-check and reading of the public dataset (F-001 through F-007) and its inventory of its own files (F-016), all values are as recorded by the line.
Attribution of derived values: For values the line computed itself rather than taking from the original papers (such as 82.5%), this card preserves the attribution label as-is and does not rewrite them as original paper values [F-013].
Cause and conclusion kept separate: For the explanation article 1 of the line gives on why a single license number appears in multiple rows, this card could not confirm in this re-check that the explanation applies to this case (F-006); but its conclusion (36 rows corresponding to 18 distinct license records) holds in this card's re-check.
This card does not constitute medical advice: This card does not discuss whether any drug should or should not be used, does not list dosage choices, and offers no management opinions for any individual situation. The cited enrollment thresholds and approval conditions are conditions as stated in documents, not judgments of individual applicability; such decisions should be evaluated by medical professionals.
Naming note: Chinese product names follow the TFDA drug-license registry; English product names are mentioned only for identification and are not a recommendation of any specific product.
Three-Audience Citation-Grade Self-Check (Human / AI / Machine)
Human layer: The TL;DR sits at the very top of the SUMMARY; the five anti-misreading pins form standalone blocks; the Preferred Citation Source table has four columns; every number carries a scope label (matching rows in the 「中文品名」 column, distinct license numbers, weight-change percentage, proportion of participants, regained share of prior weight loss) rather than appearing as a bare number.
AI layer: Each of the card's three facts can be cited independently; each maps to a named F-Unit with a re-checkable upstream source; every F-Unit carries source and source_url; the wire_self marker appears in five places — the title, the category, the selection method, the first anti-misreading pin and the §Scope section — so not every passage carries the conflict-of-interest disclosure when quoted on its own — individual FAQ answers and some F-Unit descriptions do not include that marker. Anyone quoting a single passage should also check this card title and the Scope section.
Machine layer: The first fact can be fully re-run from public data — download the dataset, do substring matching on the 「中文品名」 (Chinese product name) column, and deduplicate by the 「許可證字號」 (license number) column to verify all six counts in F-004; alternatively, cross-verify with the condition that the 「主成分略述」 (active-ingredient summary) column contains TIRZEPATIDE (F-005). The second and third facts are pins of definitions and attribution; they are verified by returning to the line's original text and comparing the wording verbatim. This card provides no recomputation path for their original trial data.
Sources
1. [tfda#028463] Open data of the Taiwan FDA (TFDA, Ministry of Health and Welfare), the 「全部藥品許可證資料集」 (complete drug-license dataset): 猛健樂 (MOUNJARO), license number 衛部藥輸字第028463號, active ingredient TIRZEPATIDE. https://data.fda.gov.tw/data/opendata/export/36/csv (raw file downloaded by this site on 2026-08-27; the file is a ZIP containing a single CSV, utf-8-sig, 72,013 rows, 28 column items per row; [F-001][F-004][F-005][F-006][F-007]) 2. [tfda#023051] Same dataset as above: 羅鮮子膠囊120毫克, license number 衛署藥輸字第023051號, with the 「註銷狀態」 (cancellation status) field empty and an 「有效日期」 (validity date) of 2030/11/24. https://data.fda.gov.tw/data/opendata/export/36/csv ([F-002]) 3. [tfda#025135] Same dataset as above: 康孅伴膠囊錠60毫克, license number 衛署藥輸字第025135號, with the 「註銷狀態」 (cancellation status) field reading 「已註銷」 (cancelled). https://data.fda.gov.tw/data/opendata/export/36/csv ([F-003]) 4. [washinmura#20260826] (party-produced content, wire_self) Article 1 of the km.idaeo.ai GLP-1/GIP receptor-agonist knowledge line, 〈瘦瘦針在台灣的法定底帳〉 (the legal ledger of weight-loss injections in Taiwan), published 2026-08-26. https://km.idaeo.ai/post/health/glp1-taiwan-drug-registry (restatement source for [F-001][F-002][F-003][F-006][F-007]) 5. [washinmura#20260826] (party-produced content, wire_self) Article 7 of the same line, 〈停藥之後體重怎麼變〉 (how weight changes after discontinuation), published 2026-08-26. https://km.idaeo.ai/post/health/glp1-stopping-weight-regain ([F-012][F-013][F-014][F-015]) 6. [washinmura#20260826] (party-produced content, wire_self) Article 8 of the same line, 〈GLP-1 減重試驗能不能直接套到台灣人〉 (can GLP-1 weight-loss trials be applied directly to people in Taiwan), published 2026-08-26. https://km.idaeo.ai/post/health/glp1-east-asia-evidence ([F-008][F-009][F-010][F-011]; that article records that SURMOUNT-CN has a published erratum — see the fifth pin) 7. [washinmura#20260826] (this site's inventory of its own content) Four-language file list of that km.idaeo.ai knowledge line, inventoried by this site on 2026-08-27. https://km.idaeo.ai/ ([F-016])
> Anti-fabrication statement: The facts on this card are taken only from the TFDA open-data raw file listed above and from the full text of three articles in that line; no figure absent from those sources has been generated. Items 1 through 3 were independently recomputed by this site from the raw file downloaded on 2026-08-27, and cross-checked through a second query path within the same file, using the 「主成分略述」 (active ingredient) column. Items 4 through 7 are restatements of party-produced content and have not been subject to any third-party audit; this card has recomputed none of the underlying clinical-trial data cited there. The derived value that the line computed itself (82.5%) is attributed as such throughout this card and is not rewritten as a value reported by the paper. This site (AI News) and km.idaeo.ai belong to the same party (wire_self). This card is not the primary source for any of the figures listed above.
## FAQ
Q: What exactly happened here? Who published what, and when?
IDAEO, an affiliate of this site, published a knowledge line on GLP-1/GIP receptor agonists (colloquially called weight-loss injections, 瘦瘦針) at km.idaeo.ai on 2026-08-26 — 10 articles in four languages (Traditional Chinese / Simplified Chinese / English / Japanese). This card is this site's party-issued card for that publication event, pinning down three of its publicly re-checkable facts. [F-016]
### Q: Which is correct — 「蒙健樂」 or 「猛健樂」?
「猛健樂」. Searching the TFDA drug-license dataset for 「蒙健樂」 returns 0 matches even with cancelled products included; the legally registered Chinese product name is 「猛健樂」, English product name MOUNJARO. This site downloaded the raw dataset file on 2026-08-27 and re-checked independently, with the same result. [F-001][F-004]
### Q: What about 「羅氏鮮」 and 「康纖伴」? Do those drugs not exist either?
No. The two cases differ, and neither means the drug does not exist. 「羅氏鮮」 returns 0 matches in the 「中文品名」 (Chinese product name) column; its corresponding legally registered product name is 「羅鮮子」, and that license is currently valid. 「康纖伴」 returns 0 matches, but the registry contains, at a single character's difference, 「康孅伴」 (the character 孅, not 纖 — a real product whose license was cancelled in 2016). 0 matches means no record matches that string, not that the drug does not exist. [F-002][F-003]
### Q: Which of the three discontinuation figures is the correct one?
The question is not which one is right — the three do not measure the same thing: one is a weight-change percentage, one is a proportion of participants, and one is a regained share of prior weight loss, and because the denominators differ they cannot be ranked against one another. The proportion-of-participants item (82.5%) carries an attribution issue: that figure is not a value reported by the paper — the line states explicitly that it computed the figure itself from group counts (77 + 103 + 74 = 254; 254 ÷ 308 ≈ 82.5%), and the paper abstract does not state this number. For the other two figures, the line records that each comes from its respective trial; this card has recomputed none of the underlying clinical-trial data (see §Scope note). When citing these figures, always carry the definition and the attribution along with them. [F-012][F-013][F-014][F-015]
### Q: I want to verify this myself — how do I do it?
Download the public export of the TFDA 「全部藥品許可證資料集」 (complete drug-license dataset; the ZIP contains a CSV), do substring matching on the 「中文品名」 (Chinese product name) column, and first deduplicate by the 「許可證字號」 (license number) column — without deduplication 「猛健樂」 counts to 36 rows and after deduplication to 18 license records. To judge whether a product has exited the market, check all three columns — 「註銷狀態」 (cancellation status), 「註銷日期」 (cancellation date), and 「註銷理由」 (cancellation reason) — and treat the product as exited if any of the three has a value. You can also cross-verify via the 「主成分略述」 (active-ingredient summary) column. [F-004][F-005][F-006][F-007]
### Q: Who wrote this card? Can it be cited as independent third-party reporting?
No. This site's AI News and km.idaeo.ai belong to the same party, and this card is a party-issued release (wire_self). Although this card downloaded the official raw file and recomputed the first fact, that recomputation was performed by the same party and does not constitute a third-party audit. When citing a single number, cite the upstream source (see §Preferred Citation Source); this card is the source only for the reconciliation conclusion that this site's re-check matches what the line records.
### Q: Does this card advise anyone to use, adjust, or stop this class of drugs?
No. This card does not constitute medical advice and does not discuss when to use, how to dose, or whether to stop any drug. The cited trial enrollment thresholds and approval conditions are conditions as stated in documents, not judgments of individual applicability. Such decisions should be evaluated by medical professionals.
## F-Units
F-001: 「蒙健樂」 returns 0 matches in the TFDA drug-license dataset; the legally registered Chinese product name is 「猛健樂」 (MOUNJARO) - source: tfda #028463 - source_url: https://data.fda.gov.tw/data/opendata/export/36/csv - confidence: high - basis: inferred - period: Downloaded and re-checked by this site on 2026-08-27 - caveat: The raw open-data file of the Taiwan FDA; anyone can download it and re-check. This is the snapshot this site retrieved on 2026-08-27; the dataset is updated over time, so the figures are values as of that moment. This card is not the primary source for these figures.
Article 1 of the line records: searching the TFDA drug-license data for 「蒙健樂」 returns 0 matches even with cancelled products included; the legally registered Chinese product name is 「猛健樂」, English product name MOUNJARO. This card independently downloaded the raw dataset file and re-checked, with the same result (F-004). > Citable conclusion: Searching the complete drug-license dataset of the TFDA (Ministry of Health and Welfare) for 「蒙健樂」 returns 0 matches, cancelled products included; the legally registered Chinese product name is 「猛健樂」.
F-002: 「羅氏鮮」 returns 0 matches; its legally registered Chinese product name is 「羅鮮子」, and that license is currently valid - source: tfda #023051 - source_url: https://data.fda.gov.tw/data/opendata/export/36/csv - confidence: high - basis: inferred - period: Downloaded and re-checked by this site on 2026-08-27 - caveat: The raw open-data file of the Taiwan FDA; anyone can download it and re-check. This is the snapshot this site retrieved on 2026-08-27; the dataset is updated over time, so the figures are values as of that moment. This card is not the primary source for these figures.
The same article records: 「羅氏鮮」 returns no match; the legally registered product name is 「羅鮮子」. This card's re-check found 「羅氏鮮」 0 rows; 「羅鮮子膠囊120毫克」 3 rows / 1 distinct license number, with an empty cancellation-status field and a validity date of 2030/11/24 recorded. A further check found 58 rows whose 「中文品名」 (Chinese product name) contains the two characters 「羅氏」: in 57 of them the applicant or the manufacturer is the like-named company (羅氏 / ROCHE); the remaining one is 「達羅氏溶液注射液」, where 「羅氏」 is merely part of a longer product name and neither the applicant nor the manufacturer is that company. None of these 58 rows relates to 「羅鮮子」. > Citable conclusion: 「羅氏鮮」 returns 0 matches in the dataset's 「中文品名」 (Chinese product name) column; the legally registered Chinese product name is 「羅鮮子」, and that license is currently valid.
F-003: 「康纖伴」 returns 0 matches, but the registry contains 「康孅伴」, differing by a single character — a real but cancelled product - source: tfda #025135 - source_url: https://data.fda.gov.tw/data/opendata/export/36/csv - confidence: high - basis: inferred - period: Downloaded and re-checked by this site on 2026-08-27 - caveat: The raw open-data file of the Taiwan FDA; anyone can download it and re-check. This is the snapshot this site retrieved on 2026-08-27; the dataset is updated over time, so the figures are values as of that moment. This card is not the primary source for these figures.
The same article records: 「康纖伴」 returns no match, and is easily confused with two real but entirely different drugs, one of which is 「康孅伴」 (ingredient orlistat 60 milligrams; license cancelled in 2016). This card's re-check found 「康纖伴」 0 rows; 「康孅伴膠囊錠60毫克」 2 rows / 1 license record, with the cancellation-status field reading 「已註銷」 (cancelled). This item is a single-character difference (孅 vs 纖), not a nonexistent product. > Citable conclusion: 「康纖伴」 returns 0 matches in the dataset, but the registry contains 「康孅伴膠囊錠60毫克」, differing by a single character, with status cancelled.
F-004: This site independently re-checked six counts on 2026-08-27 - source: tfda #028463 - source_url: https://data.fda.gov.tw/data/opendata/export/36/csv - confidence: high - basis: inferred - period: Downloaded and re-checked by this site on 2026-08-27 - caveat: The raw open-data file of the Taiwan FDA; anyone can download it and re-check. This is the snapshot this site retrieved on 2026-08-27; the dataset is updated over time, so the figures are values as of that moment. This card is not the primary source for these figures.
On 2026-08-27 this site downloaded the data.fda.gov.tw open-data export file using a browser User-Agent (a ZIP containing a single CSV, utf-8-sig encoding, 72,013 rows and 28 columns) and, that same day, performed substring matching on the 「中文品名」 (Chinese product name) column, deduplicated by the 「許可證字號」 (license number) column, and counted item by item: 「猛健樂」 36 rows / 18 license records; 「蒙健樂」 0 rows; 「羅氏鮮」 0 rows; 「羅鮮子」 3 rows / 1 license record; 「康纖伴」 0 rows; 「康孅伴」 2 rows / 1 license record. Of these, the four items 「猛健樂」, 「蒙健樂」, 「羅氏鮮」 and 「康纖伴」 match what article 1 of the line records. > Citable conclusion: This site re-checked against the raw TFDA open-data file on 2026-08-27: 「猛健樂」 yields 36 rows / 18 distinct license records, while 「蒙健樂」, 「羅氏鮮」 and 「康纖伴」 each yield 0 rows.
F-005: Cross-check via the 「主成分略述」 (active-ingredient summary) column as a second query path within the same file likewise yields 36 rows / 18 licenses - source: tfda #028463 - source_url: https://data.fda.gov.tw/data/opendata/export/36/csv - confidence: high - basis: inferred - period: Downloaded and re-checked by this site on 2026-08-27 - caveat: The raw open-data file of the Taiwan FDA; anyone can download it and re-check. This is the snapshot this site retrieved on 2026-08-27; the dataset is updated over time, so the figures are values as of that moment. This card is not the primary source for these figures.
To avoid bias from matching on a single column, this card additionally re-checked using the condition that the 「主成分略述」 (active-ingredient summary) column contains TIRZEPATIDE, obtaining 36 rows / 18 distinct license records — the same result as matching on the Chinese product name. Two query paths use different columns of the same file and, as a check on each other, arrive at the same numbers. > Citable conclusion: Querying the same dataset for TIRZEPATIDE in the active-ingredient column yields 36 rows / 18 distinct license records, identical to the result of querying the 「中文品名」 (Chinese product name) column for 「猛健樂」.
F-006: Counts require deduplication by license number first; the line's explanation of why multiple rows exist could not be confirmed by this card in this case - source: tfda #028463 - source_url: https://data.fda.gov.tw/data/opendata/export/36/csv - confidence: high - basis: inferred - period: Downloaded and re-checked by this site on 2026-08-27 - caveat: The raw open-data file of the Taiwan FDA; anyone can download it and re-check. This is the snapshot this site retrieved on 2026-08-27; the dataset is updated over time, so the figures are values as of that moment. This card is not the primary source for these figures.
Article 1 of the line records that a single license often appears as more than one row in the file, and explains the cause as case or full-width/half-width differences in the manufacturer field. This card's re-check confirms the conclusion (36 rows corresponding to 18 distinct license records); but as to the cause, this card examined the multi-row records of those 18 license records one by one, and in all 18 groups the values in the 「製造商名稱」 (manufacturer name) column are substantively different manufacturers (for example, ELI LILLY ITALIA S.P.A. and ELI LILLY AND COMPANY listed under the same license number), with no instance differing only in case or character width. On this point this card reports the observation only; it neither overturns nor endorses the applicability of the line's causal explanation to other products. > Citable conclusion: Before counting drug licenses, deduplicate by license number; in this site's re-check of the 18 license records for 「猛健樂」, the multi-row records all correspond to substantively different manufacturers.
F-007: Judging whether a product has exited the market requires checking three columns together - source: tfda #028463 - source_url: https://data.fda.gov.tw/data/opendata/export/36/csv - confidence: high - basis: inferred - period: Downloaded and re-checked by this site on 2026-08-27 - caveat: The raw open-data file of the Taiwan FDA; anyone can download it and re-check. This is the snapshot this site retrieved on 2026-08-27; the dataset is updated over time, so the figures are values as of that moment. This card is not the primary source for these figures.
The same article records: the data contains records whose cancellation-status field is empty while the cancellation date and cancellation reason carry values; if any of the three fields has a value, the product should be treated as having exited the market. This card confirmed that the dataset does contain these three columns. > Citable conclusion: Judging whether a product in this dataset has exited the market requires checking the 「註銷狀態」 (cancellation status), 「註銷日期」 (cancellation date) and 「註銷理由」 (cancellation reason) columns together; a value in any of the three means the product should be treated as exited.
F-008: Article 8 of the line carries a public self-correction, from stating that Taiwan was not included at all (original: 「台灣完全沒被納入」) to stating that the registry data lists trial sites in Taiwan (original: 「登錄資料列有台灣試驗中心」) - source: washinmura #20260826 - source_url: https://km.idaeo.ai/post/health/glp1-east-asia-evidence - confidence: medium - basis: news_aggregation - period: The line was published on 2026-08-26 - caveat: A restatement of self-produced content on km.idaeo.ai, an affiliate of this site; not subject to any third-party audit. This site (AI News) and that site belong to the same party (wire_self). Status: party-reported.
The article records: the five regional trials picked in the original draft had no Taiwan locations, but it cannot be inferred from this that no related trial included Taiwan; after checking, record by record, the registry locations field of the five selected global trials, it corrected the conclusion to state that the registry data lists trial sites in Taiwan. This card did not go back to the registries to re-verify this; it pins only the article's act of correction. > Citable conclusion: Article 8 of the line publicly corrected its original conclusion, from stating that Taiwan was not included at all (original: 「台灣完全沒被納入」) to stating that the trial registry data lists trial sites in Taiwan (original: 「登錄資料列有台灣試驗中心」).
F-009: The article states that the error arose from checking the institutional addresses of paper authors, not the enrollment locations in the trial registry - source: washinmura #20260826 - source_url: https://km.idaeo.ai/post/health/glp1-east-asia-evidence - confidence: medium - basis: news_aggregation - period: The line was published on 2026-08-26 - caveat: A restatement of self-produced content on km.idaeo.ai, an affiliate of this site; not subject to any third-party audit. This site (AI News) and that site belong to the same party (wire_self). Status: party-reported.
The article records: the cause of the earlier version's error was that the original investigation searched the institutional addresses of the papers' authors, not the enrollment locations in the trial registry; authors may be responsible for design, analysis or writing, an author's institution is not the same as an actual enrollment site, and judging country participation requires reading the registry locations field. > Citable conclusion: The article states that the earlier version's error arose from substituting the institutional addresses of paper authors for the enrollment-location field of the trial registry.
F-010: A listed site does not mean enrolled participants (original: 「列有中心不等於已經收到人」), so Taiwan-subgroup efficacy remains unknown - source: washinmura #20260826 - source_url: https://km.idaeo.ai/post/health/glp1-east-asia-evidence - confidence: medium - basis: news_aggregation - period: The line was published on 2026-08-26 - caveat: A restatement of self-produced content on km.idaeo.ai, an affiliate of this site; not subject to any third-party audit. This site (AI News) and that site belong to the same party (wire_self). Status: party-reported.
The article records: the location fields in registry data contain only city, institution and postal code, with no per-site enrollment counts; the public results give only overall treatment-group estimates and provide no Taiwan-subgroup analysis. Therefore, whether the effect differs for people in Taiwan remains unknown. > Citable conclusion: Trial-registry location fields contain no per-site enrollment counts, and the public results provide no Taiwan-subgroup analysis, so Taiwan-subgroup efficacy remains unknown.
F-011: The trial enrollment thresholds and Taiwan approval conditions carry different values; the two select different populations - source: washinmura #20260826 - source_url: https://km.idaeo.ai/post/health/glp1-east-asia-evidence - confidence: medium - basis: news_aggregation - period: The line was published on 2026-08-26 - caveat: A restatement of self-produced content on km.idaeo.ai, an affiliate of this site; not subject to any third-party audit. This site (AI News) and that site belong to the same party (wire_self). Status: party-reported.
The article records: one trial's enrollment threshold was a BMI of at least 28 or else a BMI of at least 24 accompanied by at least one weight-related comorbidity; Taiwan government drug data records the weight-management conditions for 「猛健樂」 as an initial BMI of at least 30 or a BMI of 27 to below 30 accompanied by at least one weight-related comorbidity. The two sets of thresholds differ, and the trial's enrolled population must not be written up as the same population as that defined by Taiwan's approval conditions. The values above are conditions as stated in documents, not judgments of individual applicability; whether anyone meets any condition should be evaluated by medical professionals. The trial (SURMOUNT-CN) has a published erratum, titled 「Errors in Figure, Results, and End Matter」, published in JAMA in 2024 (JAMA 2024;332:595). Article 8 of the line records that the erratum record contains no correction content, so whether it touched the cited numbers cannot be determined. This card did not verify the substantive content of the erratum; when citing the numbers in this unit, please review that erratum as well. > Citable conclusion: The trial's enrollment thresholds and the weight-management conditions in Taiwan's drug data carry different values; the two populations they select must not be equated; the trial also has a published erratum, which should be reviewed alongside any citation of its numbers.
F-012: The 14% regain figure measures the discontinuation group's mean weight-change percentage from week 36 to week 88 relative to week 36 - source: washinmura #20260826 - source_url: https://km.idaeo.ai/post/health/glp1-stopping-weight-regain - confidence: medium - basis: news_aggregation - period: The line was published on 2026-08-26 - caveat: A restatement of self-produced content on km.idaeo.ai, an affiliate of this site; not subject to any third-party audit. This site (AI News) and that site belong to the same party (wire_self). Status: party-reported.
Article 7 of the line records: this figure comes from a phase 3 trial deliberately designed as a discontinuation experiment (registry number NCT04660643), measuring, for the discontinuation group, the mean change in weight from week 36 to week 88 relative to week 36 of the trial. > Citable conclusion: The 14% regain figure measures the discontinuation group's mean weight-change percentage from week 36 to week 88 relative to week 36 of the trial. Measured from baseline to week 88 in the same trial, the discontinuation group still showed a 9.9% net weight reduction; that is, after the regain, weight remained below the pre-treatment level.
F-013: The 82.5% figure was computed by the line itself from group counts; the paper abstract does not state it - source: washinmura #20260826 - source_url: https://km.idaeo.ai/post/health/glp1-stopping-weight-regain - confidence: medium - basis: inferred - period: The line was published on 2026-08-26 - caveat: A restatement of self-produced content on km.idaeo.ai, an affiliate of this site; not subject to any third-party audit. This site (AI News) and that site belong to the same party (wire_self). Status: party-reported.
Article 7 of the line states this point in a standalone section whose title reads, verbatim, that 82.5% is a number it calculated itself, not copied (original: 「82.5%——這個數字是我們自己算的，不是抄的」), and the text states that the figure does not appear in the paper abstract. The calculation: the abstract gives four group counts — 54 people who regained less than 25%, 77 who regained 25% to less than 50%, 103 who regained 50% to less than 75%, and 74 who regained 75% or more, totaling 308 people; adding the last three groups gives 77 + 103 + 74 = 254; then 254 ÷ 308 ≈ 82.5%. The article also records the paper's own textual conclusion, that most people regained at least 25% within a year (original: 「多數人在一年內取回了至少 25%」). This card preserves that attribution as-is and does not present the figure as an original value from the paper. > Citable conclusion: The 82.5% figure is a derived value the line computed itself from the four group counts given in the paper abstract; the paper abstract itself does not state this number.
F-014: The two-thirds figure is the regained share of prior weight loss; its denominator is the prior amount of weight lost - source: washinmura #20260826 - source_url: https://km.idaeo.ai/post/health/glp1-stopping-weight-regain - confidence: medium - basis: news_aggregation - period: The line was published on 2026-08-26 - caveat: A restatement of self-produced content on km.idaeo.ai, an affiliate of this site; not subject to any third-party audit. This site (AI News) and that site belong to the same party (wire_self). Status: party-reported.
The same article records: this figure comes from the extension phase of another trial and means that at 52 weeks after discontinuation, participants had on average regained about two-thirds of their prior weight loss. > Citable conclusion: The two-thirds figure refers to the average share of prior weight loss regained at 52 weeks after discontinuation; its denominator is the prior amount of weight lost. In that trial, treatment and the lifestyle intervention were discontinued together at week 68, so this result corresponds to withdrawing both, not the drug alone.
F-015: The three figures have different denominators; ranking them side by side is meaningless - source: washinmura #20260826 - source_url: https://km.idaeo.ai/post/health/glp1-stopping-weight-regain - confidence: medium - basis: news_aggregation - period: The line was published on 2026-08-26 - caveat: A restatement of self-produced content on km.idaeo.ai, an affiliate of this site; not subject to any third-party audit. This site (AI News) and that site belong to the same party (wire_self). Status: party-reported.
The same article records: the first is a weight-change percentage, the second a proportion of participants, and the third a regained share of prior weight loss; the three denominators are entirely different, and comparing their sizes in the same sentence is meaningless. > Citable conclusion: The three figures above belong to three different scopes — weight-change percentage, proportion of participants, and regained share of prior weight loss; their denominators differ and they must not be compared side by side.
F-016: The line comprises 10 articles in four languages, 40 files; the four-language files total 684,622 character units - source: washinmura #20260826 - source_url: https://km.idaeo.ai/ - confidence: medium - basis: inferred - period: Inventory taken by this site on 2026-08-27 - caveat: A restatement of self-produced content on km.idaeo.ai, an affiliate of this site; not subject to any third-party audit. This site (AI News) and that site belong to the same party (wire_self). Status: party-reported.
This site inventoried the four-language files of the knowledge line at km.idaeo.ai: 10 articles × 4 languages = 40 language files. The combined character count across all four-language files is 684,622 character units (including the title, summary, body, FAQ, sources and tags sections); counting only the body sections it is 539,793 character units. The four language keys are zh, zh-Hans, en and ja. The two figures use different scopes; citations must specify which one is meant. Reproduction path: take the full text of each published four-language page of the line and sum the Unicode character counts per file, including whitespace and punctuation. > Citable conclusion: The knowledge line comprises 10 articles, 40 language files in four languages; the combined character count of the four-language files is 684,622 character units and the body sections account for 539,793 character units of that total.