SCALE Obesity and Prediabetes is the trial that put liraglutide 3.0 mg on the market as a weight drug, and it is almost always summarized in one line: about 8% off, roughly 5.6 kg more than placebo. That line is accurate and it leaves out the two most instructive tables in the report. More than a quarter of the drug arm and more than a third of the placebo arm never reached week 56. And buried in the stratification is a genuine randomized withdrawal, run inside the trial, that almost nobody quotes. Where this molecule sits against the newer ones is settled in the liraglutide comparison; the glycemic argument that the prediabetes half of the trial feeds is in the prediabetes article. This page is about the trial’s own construction.
What was run
A 56-week, double-blind, placebo-controlled, parallel-group, multinational trial enrolled 3,731 adults without type 2 diabetes who had a body-mass index of at least 30, or at least 27 with treated or untreated dyslipidemia or hypertension. Assignment was 2:1 to once-daily subcutaneous liraglutide 3.0 mg (2,487) or placebo (1,244), both with counseling on lifestyle modification.[1]
The structure underneath that is a four-way split, because participants were stratified by prediabetes status at randomization and the two strata ran for different lengths. The registry records 959 liraglutide and 487 placebo participants without prediabetes on a 56-week schedule, and 1,528 liraglutide and 757 placebo participants with prediabetes on a schedule running to 160 weeks of treatment.[2] Recruitment started in June 2011; the 56-week primary completion date was March 2013.[2]
Four endpoints were registered as primary, not one: the change from baseline in fasting body weight at week 56, the proportion losing at least 5% at week 56, the proportion losing more than 10% at week 56, and the proportion with onset of type 2 diabetes at 160 weeks.[2] The paper describes the first three as coprimary.[1] The fourth belongs to the prediabetes stratum alone, and what the three-year data showed is covered in the long-term article.
At baseline the cohort averaged 45.1 years (SD 12.0), 106.2 kg (SD 21.4) and a body-mass index of 38.3 (SD 6.4); 78.5% were women and 61.2% had prediabetes. Mean glycated hemoglobin was 5.6%.[1]
The headline, in two units that must not be swapped
The paper reports the primary endpoint in kilograms, with last-observation-carried-forward imputation: participants on liraglutide lost 8.4 kg (SD 7.3) and those on placebo 2.8 kg (SD 6.5), a difference of −5.6 kg (95% CI, −6.0 to −5.1), P < 0.001. A 5% reduction was reached by 63.2% against 27.1%, and a reduction of more than 10% by 33.1% against 10.6%.[1]
The registry reports the same endpoint as a percentage: −7.98% (SD 6.67) against −2.62% (SD 5.74), an estimated mean difference of −5.39 percentage points (95% CI, −5.82 to −4.95), with odds ratios of 4.80 (95% CI, 4.12 to 5.60) for the 5% threshold and 4.34 (95% CI, 3.54 to 5.32) for the 10% one.[2] Waist circumference fell 8.19 cm against 3.94.[2]
Those two difference figures — 5.6 and 5.39 — are close enough to be mistaken for each other and they are not the same quantity. One is kilograms, the other is percentage points of body weight. Any sentence carrying either has to name its unit, and neither can be converted into the other without a baseline weight it does not carry.
The ledger of who left
Over the 56-week main period, 698 of 2,487 liraglutide participants (28.1%) and 443 of 1,244 placebo participants (35.6%) did not complete. The registry posts the reasons, and they do not distribute evenly.[2]
Adverse events accounted for 238 departures from the liraglutide arm and 45 from placebo — 9.6% against 3.6%, close to a threefold difference, and the clearest asymmetry in the trial. Running the other way, lack of efficacy accounted for 23 and 36, and meeting withdrawal criteria — the largest single category on both sides — for 294 and 261, which is 11.8% against 21.0%. Protocol violations were 65 and 38, and 78 and 63 departures are filed as unclassified.[2]
This matters for how the primary estimate should be read. The published analysis carries the last observed value forward for anyone who left, which for a third of the control arm means the result rests on a weight recorded before the trial ended. Nothing about that is irregular for a 2015 obesity trial; it is simply an assumption doing more work than a reader of the headline would guess. In the prediabetes stratum the problem compounds: over the full course the registry records 722 of 1,505 liraglutide participants (48.0%) and 422 of 749 placebo participants (56.3%) not completing, with adverse events accounting for 191 and 43.[2] The three-year half of SCALE lost about half of each arm.
The randomized withdrawal nobody quotes
The participants without prediabetes did not simply finish at week 56. They were re-randomized for a 12-week off-drug period running to week 68: 351 continued liraglutide, 350 switched to placebo, and the 304 who had been on placebo all along carried on.[2] That is a preregistered randomized withdrawal, embedded in a trial from 2015, and it is the cleanest thing in the report.
Between weeks 56 and 68, weight changed +0.69% (SD 2.58) in the group kept on the drug, +2.91% (SD 3.01) in the group switched to placebo, and +0.28% (SD 2.39) in the group that had been on placebo throughout.[2] Measured from baseline instead, week 68 stood at −8.44% for continuers, −6.77% for switchers, and −3.11% for the placebo-throughout group.[2]
Twelve weeks without the drug cost roughly 2.2 percentage points of body weight relative to continuing it, and did so under randomization rather than observation. That is a small window and a single stratum, and it should not be inflated into a durability finding. It is, though, direct randomized evidence of the direction, in a trial usually cited only for its week-56 number. The broader picture of what happens after these drugs stop is in the discontinuation article.
Side effects, over unequal amounts of time
The abstract summarizes the 56-week safety picture briefly: the most frequently reported adverse events on liraglutide were mild or moderate nausea and diarrhea, and serious events occurred in 6.2% of the liraglutide group against 5.0% on placebo.[1]
The registry’s adverse-event table is a different object. It posts four groups — the prediabetes stratum followed to as long as 172 weeks, and the no-prediabetes stratum followed to 68 — so the percentages below are the share of participants who had at least one such event over their own follow-up, not 56-week rates, and the drug arm was followed longer on average than in the paper’s analysis. Pooled across all four groups: nausea in 1,040 of 2,481 (41.9%) against 206 of 1,242 (16.6%), diarrhea 589 (23.7%) against 153 (12.3%), constipation 536 (21.6%) against 129 (10.4%), vomiting 472 (19.0%) against 67 (5.4%), decreased appetite 271 (10.9%) against 43 (3.5%), fatigue 233 (9.4%) against 85 (6.8%), and headache 400 (16.1%) against 182 (14.7%).[2]
Two columns deserve separate attention. Events coded as hypoglycemia were recorded in 471 of 2,481 (19.0%) against 64 of 1,242 (5.2%) — in a trial that excluded type 2 diabetes. Those are reported adverse-event terms rather than confirmed biochemical episodes, and the trial reports no severe events in this population, but a fivefold column in non-diabetic participants is not nothing.
And the gallbladder. Serious cholelithiasis was recorded in 29 of 2,481 liraglutide participants (1.2%) against seven of 1,242 (0.6%), with serious acute cholecystitis in 13 against one, and serious pancreatitis terms in seven against one.[2] Those are small counts with no tests attached and unequal follow-up behind them, and they are the safety signal the label for this class carries. Pooled serious adverse events across the whole trial were 289 of 2,481 (11.6%) against 115 of 1,242 (9.3%) — which is the same trial as the abstract’s 6.2% against 5.0%, measured over a longer window.[2]
What SCALE did not establish
It ran no active comparator. There is no arm of orlistat, no arm of phentermine-topiramate, no arm of any injectable rival, and no supervised intensive-lifestyle arm. The head-to-head that eventually ranked this molecule against a weekly one came a decade later and is a different trial, covered in the liraglutide comparison. Nor does anything here bear on whether daily and weekly schedules produce different adherence, which is the subject of the dosing-frequency article.
It ran no cardiovascular outcome and enrolled nobody with type 2 diabetes, so nothing about cardiovascular risk can be read out of it in either direction. What this molecule’s cardiovascular evidence actually consists of is set out in the liraglutide comparison.
Its randomized withdrawal covers 12 weeks in one stratum, which is enough to show a direction and not enough to describe a year. Its 160-week primary belongs to the prediabetes stratum only, and participants who withdrew were not followed afterward, so that endpoint counts events among people still in the trial. And with 28% and 36% attrition at week 56, the precision implied by a confidence interval of −6.0 to −5.1 kg rests on an imputation rule as much as on the data.
What was in the injection, and how it compares
For scale: the weekly molecule that followed produced a mean body-weight change of −14.9% against −2.4% over 68 weeks in 1,961 adults, an estimated treatment difference of −12.4 percentage points (95% CI, −13.4 to −11.5).[3] SCALE’s −5.39 percentage points at 56 weeks is less than half of that. The two trials ran different molecules on different schedules for different lengths in different decades, so the gap is a cross-trial observation with no interval attached, but it is the right order of magnitude and it is why prescribing moved.
Every figure above belongs to branded liraglutide supplied by the sponsor, injected daily under protocol alongside lifestyle counseling, in a cohort with a mean body-mass index of 38.3 that was screened for diabetes at entry and monitored for 56 weeks or longer. Sellers listed on the compounded semaglutide board dispense compounded preparations, which are not FDA-approved and are not reviewed by the FDA for safety, efficacy or quality before they reach a patient.
The number to carry out of SCALE is not 8.4 kg. It is 5.6 kg, or equivalently 5.39 percentage points, over 56 weeks — and alongside it, the fact that 9.6% of the drug arm left because of an adverse event while 3.6% of the placebo arm did.