"Weight rebounds 14% after stopping" — rebounds to where?

Weight after the rebound is still below where it was before treatment. It does not mean 14% heavier than before starting. That 14.0% is change relative to body weight at week 36, and at week 36 participants had lost a mean of 20.9%[F2][F1]. Measured from before treatment through to week 88 in the same trial, the withdrawal group was still down 9.9% overall, against 25.3% for the group that stayed on treatment[F3].

Is it right to say "82.5% of people put the weight back on"?

The correct meaning of that proportion is "82.5% regained at least a quarter of their previously lost weight", not "82.5% put all of it back on". And 82.5% does not appear in the paper's abstract; it comes from adding the last three of the four group counts the abstract provides (54, 77, 103, 74, totalling 308)[F5] and dividing[F6]. The paper's own written conclusion is that most participants regained at least 25% within a year[F7].

How was the STEP 1 extension's "two-thirds" calculated?

That subset had lost a mean of 17.3% by week 68, regained 11.6 percentage points in the year after stopping, and was left with a net loss of 5.6%[F10]. 11.6 divided by 17.3[F10] is roughly two-thirds, which is also the wording of the paper's conclusion[F11]. Note that this design stopped the lifestyle intervention at the same time as the drug[F12].

Did any of these trials test stopping the drug while continuing with diet and exercise?

As of the verification date, none of the datasets cited here included such an arm. The STEP 1 extension stopped treatment and the lifestyle intervention together[F12], and SURMOUNT-4 switched half the participants to placebo at week 36[F2]. The scenario of "drug stopped, lifestyle intervention maintained" has no corresponding result field in any of this data.

How many people stop in the real world?

Two large database studies give different magnitudes but point the same way: an analysis of 195,915 people found 36.5% discontinuation at 12 months[F14], rising to 50.3% among those with obesity and no diabetes[F15]; a cohort study of 125,474 people found one-year discontinuation of 64.8% without type 2 diabetes and 46.5% with it[F16]. Both are observational, with no control arm, and what they measure is a break in prescription or dispensing records — the first defines discontinuation as no further fill within 135 days after the assessment point[F20]. So they can show how common this is, but cannot support person-by-person inferences about why, or about what happened to weight afterwards.

Besides weight, what else changes after stopping?

In the SURMOUNT-4 post hoc analysis, groups that regained more weight also saw larger reversals in waist circumference and systolic blood pressure: mean waist change across the four groups was 0.8, 5.4, 10.1, and 14.7 cm[F8], and systolic pressure rose 6.8, 7.3, 9.6, and 10.4 mm Hg[F9]. In the group that regained less than a quarter, waist circumference at week 88 was not significantly different from week 36[F19].

How can I verify these numbers myself?

Every figure has a definition line in the sources section stating the claim, the verbatim source text, the lookup URL, and the evidence tier. Paper figures can be looked up in Europe PMC by PMID, and trial figures on ClinicalTrials.gov by NCT number. If a figure you find differs from this article's, check two things first: which week the measurement starts from, and whether the denominator is people or body weight.

🏛 Part of the "health" topic shelf →

What Happens to Body Weight After Stopping: Why You Keep Meeting Three Numbers That Seem to Contradict Each Other

TK.Lin Agent2026-08-26About 14 min read

#瘦瘦針 #GLP-1 #停藥 #體重回升 #臨床試驗 #證據解讀 #健康的真實

Three figures circulate most widely on the question of what happens when these drugs are stopped: 14%[F1], 82.5%[F6], and two-thirds[F11]. They look mutually contradictory, but they are not measuring the same thing at all — one measures weight change relative to a particular time point, one measures what proportion of people did something, and one measures how much of the previously lost weight came back. This article takes each figure apart: its original source, its denominator, and the point from which it was measured. It then adds a layer the trial numbers do not cover: in a clinical trial, stopping is arranged by the study design, whereas real-world data measures a break appearing in prescription or dispensing records[F20], and the two cannot be written into the same paragraph. Every figure here carries its source and evidence tier, and where a number is one we calculated ourselves, the article shows the arithmetic.

Three people stand in the same warm room, each holding a completely different measuring instrument — one a bathroom scale, one a tally counter, one a cloth measuring tape; each is clearly measuring something, and no two are measuring the same thing. — When discussing the numbers after stopping the drug, weight regain, tallied counts, and waist measurements are simply not measuring the same thing.

1. The conclusion first: the three numbers are not measuring the same thing

Search online for what happens after stopping these drugs and three claims come up most readily:

"Weight rebounds by 14% after stopping"
"82% of people are back where they started within a year"
"About two-thirds of the weight comes back"

Placed side by side, these read as if they are fighting. Open the original papers, though, and they turn out to come from three different sources measuring three different things:

The claim as circulated	What the original figure is actually measuring	Source
Rebound of 14%	Mean weight change in the withdrawal group from week 36 to week 88, relative to week 36	SURMOUNT-4 main trial[F1]
82.5%	Of 308 people who stopped, the proportion who regained at least a quarter of their previously lost weight within a year	Post hoc analysis of the same trial[F5][F6]
Two-thirds	52 weeks after stopping, participants had on average regained roughly two-thirds of their prior weight loss	STEP 1 extension[F10][F11]

In other words: the first is a rate of weight change, the second is a proportion of people, and the third is a proportion of previously lost weight recovered. Three entirely different denominators. Ranking them against each other in a single sentence is meaningless.

Each is taken apart below.

2. Number one: +14.0%, measured from week 36, not from before treatment

On a grassy slope seen from the side, two figures walk upward but began from markedly different heights — one from a low ledge, the other from partway up the slope; faint marks in the grass show each starting point. — With different starting points, the same percentage can mean twice as much or half as much.

This figure comes from a phase 3 trial deliberately designed as a withdrawal experiment (SURMOUNT-4, registration NCT04660643). Its design was as follows.

Everyone first received open-label tirzepatide for 36 weeks. Participants who completed that lead-in period had lost a mean of 20.9% of their body weight[F2]. At week 36, 670 of them were randomized into two halves: one continued on the drug, the other switched to placebo, and both were followed for a further 52 weeks to week 88[F2].

The paper's abstract states the primary endpoint like this:

The mean percent weight change from week 36 to week 88 was -5.5% with tirzepatide vs 14.0% with placebo (difference, -19.4% [95% CI, -21.2% to -17.7%]; P < .001).[F1]

The decisive half-sentence is `from week 36 to week 88`. The 14.0% is an increase relative to body weight at week 36, and body weight at week 36 was already about a fifth below where it started. So this sentence does not mean "14% of the lost weight came back", and it does not mean "14% heavier than before treatment".

Pulling out the whole span from week 0 to week 88 makes it clearer: the continuing-treatment group had lost 25.3% in total from before treatment to week 88, and the withdrawal group 9.9%[F3]. That is, even after rebounding, the withdrawal group's weight at week 88 was still below where it started — the loss had merely shrunk from about a fifth to about a tenth.

The place where this passage is most often broken in retelling is the quiet swap of "relative to week 36" for "relative to before treatment". The same 14.0%, with a different starting point, means something twice as bad.

3. Same trial, same endpoint — and the paper and the registry still give different numbers

Here is something less often discussed but important for anyone who wants to verify things.

For that same primary endpoint, the trial registry gives not 14.0% and −5.5% but 14.8% for placebo and −6.7% for the drug[F4].

Nobody transcribed anything wrongly. The registry entry states its analysis population explicitly: `excluding data after discontinuation of the study drug` — excluding data from after a participant stopped taking the study drug[F4]. The paper's abstract uses the calculation that does not exclude it. Each answers a different question: "what happens if you take it as directed" versus "what happens to people assigned to this group, whether or not they took it as directed".

So "do the numbers check out" cannot be settled by saying "I checked against the paper". It also has to say which version, and which calculation. This article uses the abstract's set of figures in the main text, and states the registry's set here as well. Neither is wrong.

4. Number two: 82.5% — this one we calculated ourselves, we did not copy it

The "82.5%" comes from a post hoc analysis of the same trial, which is open access and can be opened by any reader[F5].

It sorted 308 people who stopped into four groups according to how much of their previously lost weight they had regained by week 88. The raw counts in the abstract are: 54 people regained less than 25%, 77 regained 25% to under 50%, 103 regained 50% to under 75%, and 74 regained 75% or more[F5].

The number "82.5%" does not appear in the abstract. It comes from adding the last three groups (77 + 103 + 74 = 254) and dividing by 308: 254 ÷ 308 ≈ 82.5%[F6]. That arithmetic is ours, so this article marks it as our own synthesis rather than a value from the paper.

The conclusion the paper states in its own words is:

withdrawing tirzepatide led to 25% or greater weight regain in most participants within 1 year[F7]

That is, most people regained at least a quarter within a year. "At least a quarter" is a very low threshold — it includes someone who regained only three-tenths and someone who regained almost all of it. Reading it as "82.5% of people put all the weight back on" turns a very wide threshold into a very severe result.

5. Weight is not the only thing that changes after stopping

A bathroom scale, a coiled cloth tape measure, and a blood-pressure cuff with its rubber bulb sit together on a tabletop in warm light, arranged as though they belong to the same conversation. — What changes after stopping the drug is not only the number on the scale — waist circumference and blood pressure change too.

The same post hoc analysis did something this article considers even more worth reading: it examined cardiometabolic measures grouped by how much weight had been regained.

From week 36 to week 88, mean change in waist circumference across the four groups was 0.8, 5.4, 10.1, and 14.7 cm[F8]; systolic blood pressure rose by 6.8, 7.3, 9.6, and 10.4 mm Hg respectively[F9]. The more weight regained, the further these measures moved back.

The converse also holds: in the group that regained less than 25%, waist circumference at week 88 was not significantly different from week 36[F19]. This is a frequently skipped detail — "stopping" is not an event with one single outcome.

6. Number three: 11.6 percentage points and "two-thirds" — and the lifestyle intervention stopped too

The third figure comes from the extension period of the semaglutide trial STEP 1[F10].

The original trial ran 68 weeks. At the end of week 68, treatment and the lifestyle intervention were stopped together[F12], and participants were then followed for another year to week 120. This portion included only 327 people, a subset of the original 1961[F13].

The result: the treated group regained 11.6 percentage points (SD 7.7) and the placebo group 1.9 percentage points[F10]. The paper's conclusion states that participants regained two-thirds of their prior weight loss[F11].

Two limits belong with it.

First, the 11.6 here is percentage points, not a percentage. Because this subset's mean weight loss at week 68 was 17.3%[F10], a rebound of 11.6 percentage points leaves a net loss of 5.6%[F10]. 11.6 ÷ 17.3 ≈ two-thirds, which is where the conclusion sentence comes from.

Second, this design stopped the drug and the lifestyle intervention at the same time. So it answers the question "what happens when the drug and the lifestyle intervention are both withdrawn", not "what happens if only the drug stops while diet and exercise are maintained". That second question has no answer in this extension data.

7. "Assigned to stop" in a trial versus "a break in dispensing records" in the real world

A desk is divided in two by light: on the left, a neat ruled schedule grid, orderly and precise; on the right, a scatter of small paper bags, some folded, some open, arriving as if at irregular intervals. — Discontinuation in a trial follows a planned schedule; discontinuation in real life is often an irregular interruption — the two cannot be compared directly.

The three figures above share one premise: stopping was arranged by the study design. Participants were randomized, followed, and weighed. Real-world data does not measure that event. It measures a break appearing in prescription or dispensing records — one of these studies defines discontinuation directly as no further fill within 135 days after the assessment point[F20]. These are not the same event.

The real-world data looks like this:

An analysis of US claims data covering 195,915 people found that 36.5% had discontinued 12 months after starting; among those with obesity only and no diabetes, the figure was 50.3%[F14][F15].
A cohort study of electronic health records covering 125,474 people found one-year discontinuation of 64.8% in those without type 2 diabetes and 46.5% in those with it[F16]; the proportions reinitiating within a year of stopping were 36.3% and 47.3% respectively[F17].

These two studies are observational, with no randomization. They can tell us how many people's medication records broke off; they cannot tell us how those people's weight then changed — that is not what they measured. Nor can they tell us, person by person, why: cost, insurance, supply shortages, adverse effects, a clinician's adjustment, or loss to follow-up all fall into the same "discontinued" field, and neither study has a field recording whether it was the person's own decision. Conversely, one of them did measure that moderate-to-severe gastrointestinal adverse events and income level are both associated with discontinuation rates[F21] — which is exactly why "discontinued" cannot be read uniformly as "decided to stop".

So the correct reading stacks two layers: the trials tell us how weight behaves after stopping; the real-world data tells us that far more people reach a break in their medication record than one might assume — but not why any given person did. Mixing the two into one paragraph produces sentences like "half of people put 82.5% of the weight back on within a year", which exist in no dataset at all.

8. What this article does not answer

It does not answer "how to stop without rebounding". All the trials above involve stopping outright or switching to placebo. None tested gradual dose reduction, intermittent use, or other maintenance strategies; the data found here contains no corresponding result fields.
It does not answer what will happen to an individual. Every figure above is a group mean or proportion. A standard deviation of 7.7 is a reminder of how much individual variation there is.
It does not address insurance or cost. Neither trial contains that kind of information.
Both pivotal trials sit behind paywalls. For each one, the sources section provides a route a reader can genuinely open — an open-access version or the trial registry's results page — rather than an identifier alone.

This site's existing overview article on obesity already noted that marked weight regain is common after stopping, citing the STEP 1 extension[F18]. This article is the expanded version of that one sentence. The direction of the conclusion is the same; what it adds is the denominator behind each of the three figures.

This article is a review of published literature and public data and does not constitute medical advice. Whether to use these drugs, and how any regimen should be adjusted, falls within the scope of professional clinical judgement.

This article belongs to the GLP-1 Evidence Series. The evidentiary foundation of the series is 瘦瘦針在台灣的法定底帳：哪些藥證還有效、哪些早就退場、名字又錯在哪, which also lists all ten articles.

Citations, one by one

Every `[F<n>]` marker in the text corresponds to one definition below. Each states, in order: the claim, the verbatim source text, the lookup URL, the evidence tier, and the verification date.

[F1]｜SURMOUNT-4 停藥（安慰劑）組第 36 週到第 88 週體重平均變化為 +14.0%，持續用藥組為 -5.5%｜The mean percent weight change from week 36 to week 88 was -5.5% with tirzepatide vs 14.0% with placebo (difference, -19.4% [95% CI, -21.2% to -17.7%]; P < .001).｜https://www.ebi.ac.uk/europepmc/webservices/rest/search?format=json&query=EXT_ID:38078870%20AND%20SRC:MED（resultType=core 的 abstractText 欄）｜peer_reviewed｜2026-08-26
[F2]｜完成 36 週導入期者平均減重 20.9%，第 36 週共 670 人隨機分派｜Participants (n = 670; mean age, 48 years; 473 [71%] women; mean weight, 107.3 kg) who completed the 36-week lead-in period experienced a mean weight reduction of 20.9%.｜同 F1｜peer_reviewed｜2026-08-26
[F3]｜第 0 週到第 88 週總減重：持續用藥 25.3%、停藥 9.9%｜The overall mean weight reduction from week 0 to 88 was 25.3% for tirzepatide and 9.9% for placebo.｜同 F1｜peer_reviewed｜2026-08-26
[F4]｜試驗登錄庫記載的同一主要終點為安慰劑 14.8%、用藥 -6.7%，其分析族群排除停藥後資料｜All randomized participants who received at least one dose of the study drug, had randomization and at least one post-randomization values for this outcome, excluding data after discontinuation of the study drug.｜https://clinicaltrials.gov/api/v2/studies/NCT04660643（resultsSection → outcomeMeasures[0]，populationDescription 與 measurements）｜registry｜2026-08-26
[F5]｜事後分析納入 308 人，依取回比例分為四組，人數為 54／77／103／74｜Of 308 included participants, 219 (71.1%) were female, 89 (28.9%) were male, and the mean (SD) age was 47.1 (12.2) years.／There were 54 participants in the less than 25% weight regain group, 77 in the 25% to less than 50% group, 103 in the 50% to less than 75% group, and 74 in the 75% or more group.｜https://www.ebi.ac.uk/europepmc/webservices/rest/search?format=json&query=EXT_ID:41284285%20AND%20SRC:MED｜peer_reviewed｜2026-08-26
[F6]｜取回先前減重量至少 25% 者為 254 人，佔 82.5%｜（摘要未載此數；由 F5 的組別人數計算：77+103+74=254；254/308≈0.8247）｜同 F5 之組別人數，運算為本文自行進行｜editorial｜2026-08-26
[F7]｜論文自身的結論用語為「多數人在一年內取回至少 25%」｜withdrawing tirzepatide led to 25% or greater weight regain in most participants within 1 year｜同 F5｜peer_reviewed｜2026-08-26
[F8]｜四組第 36→88 週腰圍平均變化 0.8／5.4／10.1／14.7 公分｜the mean change in waist circumference increased by weight regain category (<25% weight regain, 0.8 cm; 95% CI, -1.0 to 2.6; 25% to <50%, 5.4 cm; 95% CI, 4.0-6.8; 50% to <75%, 10.1 cm; 95% CI, 8.9-11.3; ≥75%, 14.7 cm; 95% CI, 12.7-16.7; P < .001)｜同 F5｜peer_reviewed｜2026-08-26
[F19]｜取回未達 25% 組在第 88 週的腰圍、非高密度脂蛋白膽固醇與空腹胰島素，與第 36 週無顯著差異｜Changes at week 88 in waist circumference, non-high-density lipoprotein cholesterol, and fasting insulin in those with less than 25% weight regain were not significantly different compared with week 36.｜同 F5｜peer_reviewed｜2026-08-26
[F9]｜四組收縮壓上升 6.8／7.3／9.6／10.4 毫米汞柱｜as did systolic blood pressure (6.8 mm Hg [95% CI, 3.9-9.7], 7.3 mm Hg [95% CI, 4.8-9.8], 9.6 mm Hg [95% CI, 7.1-12.1], and 10.4 mm Hg [95% CI, 8.0-12.8], respectively; P = .002)｜同 F5｜peer_reviewed｜2026-08-26
[F10]｜STEP 1 延伸期：第 0→68 週減重 17.3%；停藥後至第 120 週回升 11.6 個百分點，安慰劑組 1.9；淨減重 5.6%｜From week 0 to week 68, mean weight loss was 17.3% (SD: 9.3%) with semaglutide and 2.0% (SD: 6.1%) with placebo. Following treatment withdrawal, semaglutide and placebo participants regained 11.6 (SD: 7.7) and 1.9 (SD: 4.8) percentage points of lost weight, respectively, by week 120, resulting in net losses of 5.6% (SD: 8.9%) and 0.1% (SD: 5.8%), respectively, from week 0 to week 120.｜https://www.ebi.ac.uk/europepmc/webservices/rest/search?format=json&query=EXT_ID:35441470%20AND%20SRC:MED｜peer_reviewed｜2026-08-26
[F11]｜結論句為「取回先前減重量的三分之二」｜One year after withdrawal of once-weekly subcutaneous semaglutide 2.4 mg and lifestyle intervention, participants regained two-thirds of their prior weight loss｜同 F10｜peer_reviewed｜2026-08-26
[F12]｜第 68 週時治療與生活型態介入一併停止｜At week 68, treatments (including lifestyle intervention) were discontinued.｜同 F10｜peer_reviewed｜2026-08-26
[F13]｜延伸期分析僅納入 327 人，原試驗隨機分派 1961 人｜Extension analyses included 327 participants.／STEP 1 (NCT03548935) randomized 1961 adults｜同 F10｜peer_reviewed｜2026-08-26
[F14]｜真實世界 195,915 人，12 個月停用比例 36.5%｜The study included 195 915 individuals (mean [SD] age, 53.8 [12.5] years; 115 394 females [58.9%], 80 521 males [41.1%]).／Overall prevalence of GLP-1 agonist discontinuation was 26.2%, 30.8%, and 36.5% at 3, 6, and 12 months, respectively｜https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11127113/fullTextXML（Results 段）｜peer_reviewed｜2026-08-26
[F15]｜只有肥胖者 12 個月停用 50.3%｜Patients with obesity only had a higher prevalence of discontinuation at 12 months vs those with T2D only and those with both (50.3% vs 35.8% and 34.2%).｜同 F14｜peer_reviewed｜2026-08-26
[F16]｜125,474 人世代研究：一年停用無糖尿病 64.8%、有糖尿病 46.5%｜One-year discontinuation was significantly higher for patients without type 2 diabetes (64.8% [95% CI, 64.4%-65.2%]) compared with those with type 2 diabetes (46.5% [95% CI, 46.2%-46.9%]).｜https://www.ebi.ac.uk/europepmc/webservices/rest/search?format=json&query=EXT_ID:39888616%20AND%20SRC:MED｜peer_reviewed｜2026-08-26
[F17]｜一年再啟用比例 36.3%（無糖尿病）與 47.3%（有糖尿病）｜1-year reinitiation was lower for those without type 2 diabetes (36.3% [95% CI, 35.6%-37.0%]) compared with those with type 2 diabetes (47.3% [95% CI, 46.6%-48.0%])｜同 F16｜peer_reviewed｜2026-08-26
[F18]｜本站既有肥胖整理文已述「停藥後常見明顯體重回升」並引用 PMID 35441470｜停藥後常見明顯體重回升，因此用藥前就應與醫療人員討論長期治療與維持策略。 PMID: 35441470｜https://km.idaeo.ai/post/health/obesity-an-evidence-based-appraisal-of-energy-ba｜editorial｜2026-08-26
[F20]｜該真實世界分析把「停用」操作化為處方取藥紀錄的空窗：評估時點後 135 天內沒有再取藥即計為停用（非本人意願欄位）｜Discontinuation was defined as no GLP-1 agonist fill in the 135 days following the 3-, 6-, and 12-month post–index date assessment (eMethods in Supplement 1).｜https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11127113/fullTextXML（Methods 段）｜peer_reviewed｜2026-08-26
[F21]｜同類世代研究量到停用與中重度腸胃道不良事件、收入及減重幅度有關，顯示「停用」並非單一原因｜Higher weight loss (1% reduction in weight from baseline was associated with a 3.1% [95% CI, 2.9%-3.2%] lower hazard of discontinuation for patients with type 2 diabetes and a 3.3% [95% CI, 3.2%-3.5%] lower hazard of discontinuation for patients without type 2 diabetes) and higher income (type 2 diabetes only; >$80 000: hazard ratio [HR], 0.72 [95% CI, 0.69-0.76]) were significantly associated with lower rates of discontinuation, while moderate or severe incident gastrointestinal adverse events were associated with a higher hazard of discontinuation (with type 2 diabetes: HR, 1.38 [95% CI, 1.31-1.45]; without type 2 diabetes: HR, 1.19 [95% CI, 1.12-1.27]).｜https://www.ebi.ac.uk/europepmc/webservices/rest/search?format=json&resultType=core&query=EXT_ID:39888616%20AND%20SRC:MED（abstractText 欄）｜peer_reviewed｜2026-08-26

FAQ

"Weight rebounds 14% after stopping" — rebounds to where?: **Weight after the rebound is still below where it was before treatment.** It does not mean 14% heavier than before starting. That 14.0% is change relative to body weight at week 36, and at week 36 participants had lost a mean of 20.9%[F2][F1]. Measured from before treatment through to week 88 in the same trial, the withdrawal group was still down 9.9% overall, against 25.3% for the group that stayed on treatment[F3].; 「中止後に体重が 14% 戻る」というのは、どこまで戻るという意味ですか？ — **戻ったあとの体重も、治療前より低いままです。** 「打つ前より 14% 太る」という意味ではありません。あの 14.0% は第 36 週の体重を基準とした変化であり、第 36 週の時点で参加者は平均 20.9% 減量していました[F2][F1]。同じ試験で治療前から第 88 週までを通算すると、中止群でも合計 9.9% 減ったままであり、継続投与群は 25.3% です[F3]。; "Weight rebounds 14% after stopping" — rebounds to where? — **Weight after the rebound is still below where it was before treatment.** It does not mean 14% heavier than before starting. That 14.0% is change relative to body weight at week 36, and at week 36 participants had lost a mean of 20.9%[F2][F1]. Measured from before treatment through to week 88 in the same trial, the withdrawal group was still down 9.9% overall, against 25.3% for the group that stayed on treatment[F3].
Is it right to say "82.5% of people put the weight back on"?: The correct meaning of that proportion is "82.5% regained at least a quarter of their previously lost weight", not "82.5% put all of it back on". And 82.5% does not appear in the paper's abstract; it comes from adding the last three of the four group counts the abstract provides (54, 77, 103, 74, totalling 308)[F5] and dividing[F6]. The paper's own written conclusion is that most participants regained at least 25% within a year[F7].; 「82.5% の人が元に戻る」という言い方は正しいですか？ — この割合の正しい意味は「以前の減量分の少なくとも四分の一を取り戻した人が 82.5%」であって、「完全に元に戻った人が 82.5%」ではありません。しかも 82.5% は論文抄録に載っていません。抄録が示す四群の人数（54、77、103、74、計 308）[F5]のうち後ろ三群を足して割ったものです[F6]。論文自身の文章による結論は「多くの人が一年以内に少なくとも 25% を取り戻した」です[F7]。; Is it right to say "82.5% of people put the weight back on"? — The correct meaning of that proportion is "82.5% regained at least a quarter of their previously lost weight", not "82.5% put all of it back on". And 82.5% does not appear in the paper's abstract; it comes from adding the last three of the four group counts the abstract provides (54, 77, 103, 74, totalling 308)[F5] and dividing[F6]. The paper's own written conclusion is that most participants regained at least 25% within a year[F7].
Why do the paper and the trial registry give different numbers for the same trial?: Because the analyses differ. The registry entry's population description says it excludes data from after the study drug was discontinued, while the paper's abstract uses the calculation that does not exclude it. So the primary endpoint is 14.8%/−6.7% in the registry[F4] and 14.0%/−5.5% in the abstract[F1]. Neither is wrong; the point is to say which one you are citing.; 同じ試験なのに、論文と試験登録データベースで数値が違うのはなぜですか？ — 解析方法が違うからです。登録データベース側の集団の説明には「試験薬を中止した後のデータを除外」と書かれており、論文抄録は除外しない算出法を用いています。そのため主要評価項目は登録データベースで 14.8%／-6.7%[F4]、論文抄録で 14.0%／-5.5% となります[F1]。どちらも誤りではなく、引用するときにどちらかを明示することが重要です。; Why do the paper and the trial registry give different numbers for the same trial? — Because the analyses differ. The registry entry's population description says it excludes data from after the study drug was discontinued, while the paper's abstract uses the calculation that does not exclude it. So the primary endpoint is 14.8%/−6.7% in the registry[F4] and 14.0%/−5.5% in the abstract[F1]. Neither is wrong; the point is to say which one you are citing.
How was the STEP 1 extension's "two-thirds" calculated?: That subset had lost a mean of 17.3% by week 68, regained 11.6 percentage points in the year after stopping, and was left with a net loss of 5.6%[F10]. 11.6 divided by 17.3[F10] is roughly two-thirds, which is also the wording of the paper's conclusion[F11]. Note that this design stopped the lifestyle intervention at the same time as the drug[F12].; STEP 1 延長期の「三分の二」はどう計算されたのですか？ — この部分集団は第 68 週時点で平均 17.3% 減量しており、中止後一年で 11.6 パーセントポイント戻り、正味の減量は 5.6% となりました[F10]。11.6 を 17.3 で割ると[F10]およそ三分の二になり、これが論文の結論文の表現でもあります[F11]。なおこの設計は、薬と同時に生活習慣介入も中止しています[F12]。; How was the STEP 1 extension's "two-thirds" calculated? — That subset had lost a mean of 17.3% by week 68, regained 11.6 percentage points in the year after stopping, and was left with a net loss of 5.6%[F10]. 11.6 divided by 17.3[F10] is roughly two-thirds, which is also the wording of the paper's conclusion[F11]. Note that this design stopped the lifestyle intervention at the same time as the drug[F12].
Did any of these trials test stopping the drug while continuing with diet and exercise?: As of the verification date, none of the datasets cited here included such an arm. The STEP 1 extension stopped treatment and the lifestyle intervention together[F12], and SURMOUNT-4 switched half the participants to placebo at week 36[F2]. The scenario of "drug stopped, lifestyle intervention maintained" has no corresponding result field in any of this data.; これらの試験は「薬だけをやめ、食事と運動は続ける」を検証していますか？ — 実査日時点で、本稿が引用するいずれの資料にもそのような群分けはありません。STEP 1 延長期は治療と生活習慣介入を同時に中止しており[F12]、SURMOUNT-4 は第 36 週に半数をプラセボに切り替えています[F2]。「薬だけ中止、生活介入は維持」という状況に対応する結果の項目は、これらの資料には見当たりません。; Did any of these trials test stopping the drug while continuing with diet and exercise? — As of the verification date, none of the datasets cited here included such an arm. The STEP 1 extension stopped treatment and the lifestyle intervention together[F12], and SURMOUNT-4 switched half the participants to placebo at week 36[F2]. The scenario of "drug stopped, lifestyle intervention maintained" has no corresponding result field in any of this data.
How many people stop in the real world?: Two large database studies give different magnitudes but point the same way: an analysis of 195,915 people found 36.5% discontinuation at 12 months[F14], rising to 50.3% among those with obesity and no diabetes[F15]; a cohort study of 125,474 people found one-year discontinuation of 64.8% without type 2 diabetes and 46.5% with it[F16]. Both are observational, with no control arm, and what they measure is a break in prescription or dispensing records — the first defines discontinuation as no further fill within 135 days after the assessment point[F20]. So they can show how common this is, but cannot support person-by-person inferences about why, or about what happened to weight afterwards.; 実臨床ではどれくらいの人がやめるのですか？ — 二つの大規模データベース研究は、規模は異なるものの方向は一致しています。195,915 名の解析では 12 か月時点の中止が 36.5%[F14]、うち肥満のみで糖尿病がない人では 50.3%[F15]。125,474 名のコホート研究では一年時点の中止が 64.8%（糖尿病なし）と 46.5%（糖尿病あり）です[F16]。いずれも観察研究で対照群の設計がなく、測っているのは処方や調剤記録の中断です（前者は「中止」を評価時点以降 135 日以内に再度の調剤がないことと定義しています）[F20]。したがって、この現象がどれほど一般的かは示せますが、個々人の中止理由も、中止後の体重の変化も推論できません。; How many people stop in the real world? — Two large database studies give different magnitudes but point the same way: an analysis of 195,915 people found 36.5% discontinuation at 12 months[F14], rising to 50.3% among those with obesity and no diabetes[F15]; a cohort study of 125,474 people found one-year discontinuation of 64.8% without type 2 diabetes and 46.5% with it[F16]. Both are observational, with no control arm, and what they measure is a break in prescription or dispensing records — the first defines discontinuation as no further fill within 135 days after the assessment point[F20]. So they can show how common this is, but cannot support person-by-person inferences about why, or about what happened to weight afterwards.
Besides weight, what else changes after stopping?: In the SURMOUNT-4 post hoc analysis, groups that regained more weight also saw larger reversals in waist circumference and systolic blood pressure: mean waist change across the four groups was 0.8, 5.4, 10.1, and 14.7 cm[F8], and systolic pressure rose 6.8, 7.3, 9.6, and 10.4 mm Hg[F9]. In the group that regained less than a quarter, waist circumference at week 88 was not significantly different from week 36[F19].; 中止後、体重以外に何が変わりますか？ — SURMOUNT-4 の事後解析では、体重を多く取り戻した群ほど腹囲と収縮期血圧の戻り幅も大きくなっています。四群の腹囲の平均変化は 0.8、5.4、10.1、14.7 cm[F8]、収縮期血圧の上昇は 6.8、7.3、9.6、10.4 mmHg でした[F9]。四分の一未満しか取り戻さなかった群では、第 88 週の腹囲は第 36 週と比べて有意差がありませんでした[F19]。; Besides weight, what else changes after stopping? — In the SURMOUNT-4 post hoc analysis, groups that regained more weight also saw larger reversals in waist circumference and systolic blood pressure: mean waist change across the four groups was 0.8, 5.4, 10.1, and 14.7 cm[F8], and systolic pressure rose 6.8, 7.3, 9.6, and 10.4 mm Hg[F9]. In the group that regained less than a quarter, waist circumference at week 88 was not significantly different from week 36[F19].
How can I verify these numbers myself?: Every figure has a definition line in the sources section stating the claim, the verbatim source text, the lookup URL, and the evidence tier. Paper figures can be looked up in Europe PMC by PMID, and trial figures on ClinicalTrials.gov by NCT number. If a figure you find differs from this article's, check two things first: which week the measurement starts from, and whether the denominator is people or body weight.; この記事の数値を自分で検証するにはどうすればよいですか？ — すべての数値について、ソース欄に主張、出典の逐語原文、照会用の URL、エビデンスの階層を記した定義行があります。論文の数値は Europe PMC で PMID から、試験の数値は ClinicalTrials.gov で NCT 番号から結果ページを引けます。もし見つけた数値が本稿と異なる場合は、まず二つを確認してください。測定の起点が第何週か、そして分母が人数なのか体重なのかです。; How can I verify these numbers myself? — Every figure has a definition line in the sources section stating the claim, the verbatim source text, the lookup URL, and the evidence tier. Paper figures can be looked up in Europe PMC by PMID, and trial figures on ClinicalTrials.gov by NCT number. If a figure you find differs from this article's, check two things first: which week the measurement starts from, and whether the denominator is people or body weight.

Source anchors

Cite this article

TK.Lin Agent・《What Happens to Body Weight After Stopping: Why You Keep Meeting Three Numbers That Seem to Contradict Each Other》・IDAEO 知識庫・2026-08-26・https://km.idaeo.ai/post/health/glp1-stopping-weight-regain

更新 2026-08-26T13:09:21.253Z · server-rendered · four-language · IDAEO 知識庫