GLP-1 Agonist Weight Loss Trials: 68-Week Data by Drug

The largest placebo-controlled GLP-1 agonist weight loss clinical trial data comes from the semaglutide STEP program. In STEP 3, 611 adults were randomized 2:1 to semaglutide 2.4 mg subcutaneously once weekly or placebo, both alongside intensive behavioral therapy, and mean body weight change at week 68 was -16.0% against -5.7%, a difference of 10.3 percentage points (95% CI -12.0 to -8.6, P<.001) (STEP 3, PMID 33625476). Liraglutide 3.0 mg daily reached -7.8% at 52 weeks in the phase 2 trial that ran it against semaglutide and placebo (NCT02453711, PMID 30122305).

How much weight did each GLP-1 agonist lose in its own trial?

Trial (PMID)	Drug, dose, route	n randomized	Duration	Active arm	Comparator arm	Difference
STEP 3 (PMID 33625476)	semaglutide 2.4 mg SC weekly + intensive behavioral therapy	611 (407 / 204)	68 weeks	-16.0%	-5.7% placebo	-10.3 pp (95% CI -12.0 to -8.6)
STEP 4 (PMID 33755728)	semaglutide 2.4 mg SC weekly, continued after a 20-week run-in	803 (535 / 268)	week 20 to week 68	-7.9%	+6.9% switched to placebo	-14.8 pp (95% CI -16.0 to -13.5)
NCT02453711 phase 2 (PMID 30122305)	semaglutide 0.4 mg SC daily	957 total, 102-103 per active arm	52 weeks	-13.8%	-2.3% placebo	-11.5 pp (computed: 13.8 - 2.3)
NCT02453711 phase 2 (PMID 30122305)	liraglutide 3.0 mg SC daily	957 total, 102-103 per active arm	52 weeks	-7.8%	-2.3% placebo	-5.5 pp (computed: 7.8 - 2.3)
Copenhagen / Hvidovre (PMID 38916894)	liraglutide 3.0 mg SC daily after an 8-week 800 kcal/day diet	195 (49 liraglutide / 49 placebo)	52 weeks	-13.74 kg (95% CI 11.04 to 16.44)	-7.03 kg (95% CI 4.25 to 9.80)	-6.71 kg (computed: 13.74 - 7.03)

Table 1: primary weight endpoints by trial, with within-trial differences either as reported or computed by subtraction where shown. Sources: PMID 33625476, PMID 33755728, PMID 30122305, PMID 38916894.

Two rows in that table use percentages and one uses kilograms, and they are not interchangeable from the published records. The Copenhagen trial reports mean BMI 37.00 (SD 2.92) but no mean body weight, so its 13.74 kg cannot be converted to a percentage without inventing a denominator (PMID 38916894). STEP 3 supplies both: a mean baseline weight of 105.8 kg (SD 22.9) against -16.0% gives 105.8 x 0.160 = 16.93 kg, and the placebo arm’s -5.7% gives 105.8 x 0.057 = 6.03 kg, for a 10.90 kg gap (PMID 33625476).

Dose is the other axis the headline numbers hide. The phase 2 trial escalated semaglutide from 0.05 mg per day in 4-week steps and held five separate maintenance doses to week 52, which is the only published GLP-1 agonist weight loss clinical trial data set with a full daily dose-response curve in adults without diabetes.

Daily SC dose, 52 weeks	Weight change	Placebo-adjusted (computed, minus -2.3%)
Placebo	-2.3%	0 pp
Semaglutide 0.05 mg	-6.0%	-3.7 pp
Semaglutide 0.1 mg	-8.6%	-6.3 pp
Semaglutide 0.2 mg	-11.6%	-9.3 pp
Semaglutide 0.3 mg	-11.2%	-8.9 pp
Semaglutide 0.4 mg	-13.8%	-11.5 pp
Liraglutide 3.0 mg	-7.8%	-5.5 pp

Table 2: estimated mean weight loss at week 52 by arm, with the placebo-adjusted column computed as (arm % - 2.3%). All figures from PMID 30122305; bodyweight data were available for 891 of 957 participants (93%) at week 52.

The curve is not monotonic. Semaglutide 0.3 mg produced -11.2% against -11.6% at 0.2 mg, so the fourth dose step landed below the third (PMID 30122305). Across the full 8-fold dose range, 0.4 mg versus 0.05 mg (0.4 / 0.05 = 8), the placebo-adjusted effect grew only 3.1-fold (11.5 / 3.7 = 3.1), which is the shape you would expect from a saturating receptor response rather than a linear one.

Does semaglutide weight loss hold if the drug is stopped?

STEP 4 was built as a withdrawal trial, and its comparator arm moves in the opposite direction from every other placebo group here. After a 20-week run-in in which 902 participants received semaglutide and 803 (89.0%) reached the 2.4 mg weekly maintenance dose, randomization was 2:1 to continue (n=535) or switch to placebo (n=268) for 48 more weeks (PMID 33755728). The 99 who did not reach maintenance dose represent 902 - 803 = 99, or 11.0% of the run-in cohort.

From week 20 to week 68 the continued arm changed -7.9% and the switched arm +6.9% (difference -14.8 pp, 95% CI -16.0 to -13.5, P<.001), against a mean randomization weight of 107.2 kg (SD 22.7) (PMID 33755728). In kilograms that is 107.2 x 0.079 = 8.47 kg lost and 107.2 x 0.069 = 7.40 kg regained.

Chaining the two phases of that one trial gives the total from pre-run-in weight. The run-in mean loss was 10.6%, so continued treatment ends at 1 - (0.894 x 0.921) = 1 - 0.8234 = 17.7% below starting weight, while withdrawal ends at 1 - (0.894 x 1.069) = 1 - 0.9557 = 4.4% below it (formula: 1 minus the product of the two retained-weight fractions; inputs from PMID 33755728).

Retention was high enough that the estimate is not carried by dropouts: 787 of 803 (98.0%) completed the trial and 741 (92.3%) completed treatment (PMID 33755728).

Semaglutide vs liraglutide vs tirzepatide: what do the direct comparisons show?

The phase 2 trial is the head-to-head in this set. Semaglutide at 0.2 mg per day or higher beat liraglutide 3.0 mg within the same randomization, at -13.8% to -11.2% against -7.8%, and the 0.4 mg arm exceeded liraglutide by 13.8 - 7.8 = 6.0 pp (PMID 30122305). Categorical response separated the same way: 10% or more weight loss occurred in 10% of placebo participants against 37% to 65% of those on 0.1 mg or more of semaglutide (P<0.0001 versus placebo).

Tirzepatide enters the evidence indirectly. A network meta-analysis of 28 trials and 23,622 adults with type 2 diabetes (44.2% female), searched to 11 November 2023, reported weight reductions versus placebo ranging from 9.57 kg at tirzepatide 15 mg down to 5.27 kg at 5 mg, and from 4.97 kg at semaglutide 2.0 mg down to 2.52 kg at 0.5 mg (PMID 38613667). Inside that single network rather than a head-to-head trial, the gap between tirzepatide 15 mg and semaglutide 2.0 mg is 9.57 - 4.97 = 4.60 kg.

Two features of that analysis limit how far the figure travels. The population is adults with type 2 diabetes, not the obesity-without-diabetes population of STEP 3 and the phase 2 trial, and the semaglutide doses tested top out at 2.0 mg weekly rather than the 2.4 mg used for weight management (PMID 38613667). All the compared arms are subcutaneous, which removes the route asymmetry that complicates peptide and small-molecule evidence comparisons.

On glycemia the same network ranked tirzepatide 15 mg first at -21.61 mmol/mol (-1.96%) HbA1c versus placebo, with tirzepatide 5 mg at -17.60 mmol/mol (-1.60%) landing essentially level with semaglutide 2.0 mg at -17.74 mmol/mol (-1.59%) (PMID 38613667).

Why do the placebo arms differ so much between GLP-1 trials?

Placebo arm design sets the size of the reported difference as much as the drug does. STEP 3 paired placebo with 30 counseling visits and an initial 8-week low-calorie diet, and that arm lost 5.7%; the phase 2 trial paired placebo with dietary and physical activity counselling only, and that arm lost 2.3% (PMID 33625476; PMID 30122305). Comparing two separate trials with different lifestyle programs rather than a head-to-head, the placebo spread alone is 5.7 - 2.3 = 3.4 pp.

That is why STEP 3’s active arm can lose more weight than STEP 4’s while showing a smaller between-group difference: 16.0% against a losing comparator gives -10.3 pp, whereas 7.9% against a regaining comparator gives -14.8 pp (PMID 33625476; PMID 33755728).

Responder thresholds are less sensitive to the framing, which is why they are the more portable number.

Weight loss threshold at week 68	Semaglutide 2.4 mg + IBT	Placebo + IBT	Gap (computed)	Number needed to treat (computed)
5% or more	86.6%	47.6%	39.0 pp	2.6
10% or more	75.3%	27.0%	48.3 pp	2.1
15% or more	55.8%	13.2%	42.6 pp	2.3

Table 3: categorical responder rates in STEP 3 (n=611, 68 weeks, all P<.001), with the gap computed by subtraction and NNT as 100 divided by the gap in percentage points. Source: PMID 33625476.

The widest separation sits at the 10% threshold, not at 5% or 15%. Below it the placebo arm’s own 47.6% response compresses the gap, and above it the active arm’s response rate falls to 55.8%.

Gastrointestinal events track the same trial-design dependency. STEP 3 recorded them in 82.8% of semaglutide participants against 63.2% on placebo, a 19.6 pp gap, with 3.4% of semaglutide participants and 0% of placebo participants discontinuing because of them (PMID 33625476). STEP 4 recorded 49.1% against 26.1%, a 23.0 pp gap, with discontinuation for adverse events at 2.4% and 2.2% (PMID 33755728). The absolute rates differ by more than 30 points across those two trials, which share a drug and dose but not a comparator regimen.

What did the trials record besides body weight?

STEP 4 reported waist circumference -9.7 cm (95% CI -10.9 to -8.5), systolic blood pressure -3.9 mm Hg (95% CI -5.8 to -2.0), and SF-36 physical functioning +2.5 points (95% CI 1.6 to 3.3) for continued semaglutide against the switch to placebo, all P<.001 (PMID 33755728).

Bone mineral density is where a weight-matched comparison changes the reading. The Copenhagen trial randomized 195 adults aged 18 to 65 with BMI 32 to 43 and no diabetes to exercise alone, liraglutide 3.0 mg daily, both, or placebo for 52 weeks after an 8-week 800 kcal/day diet (PMID 38916894). Liraglutide alone lost 13.74 kg against 11.19 kg for exercise alone, or 2.55 kg more, yet BMD fell further on liraglutide at both the hip (-0.013 g/cm2, 95% CI -0.024 to -0.001, P=.03) and the lumbar spine (-0.016 g/cm2, 95% CI -0.032 to -0.001, P=.04).

The combination arm lost the most weight of the four, 16.88 kg (95% CI 14.23 to 19.54), and was the arm whose BMD held: unchanged against placebo at the hip (-0.006 g/cm2, 95% CI -0.017 to 0.004, P=.24) and lumbar spine (-0.010 g/cm2, 95% CI -0.025 to 0.005, P=.20), measured by dual-energy x-ray absorptiometry in the intention-to-treat population. That trial ran from August 2016 to November 2019 and is registered as EudraCT 2015-005585-32.