Not Apples to Apples: How Shifting Trial Criteria Distort GLP-1 Weight-Loss Headlines

When investors compare weight-loss drug trials - Novo Nordisk's liraglutide (Saxenda) at 8%, semaglutide (Wegovy) at 15-21%, and Eli Lilly's tirzepatide (Zepbound) at 22% - it looks like a clean march of pharmacological progress. That framing is not wrong, but it is incomplete. The patient populations enrolled in these trials have changed significantly over the same decade. You are not just comparing drugs - you are comparing different people, subjected to different screening thresholds, followed for different lengths of time.

The Inclusion Criteria Scorecard

The table below compares the pivotal non-diabetic weight-loss trials for each approved or late-stage drug, restricting to phase 3 (or large phase 2) studies in adults without type 2 diabetes - the closest thing to a like-for-like baseline. Data sourced from ClinicalTrials.gov and published results.

Three Shifts That Compound Each Other

1. From diagnosis-based to measurement-based diabetes screening

SCALE Obesity excluded type 1 and type 2 diabetes by diagnosis. There was no HbA1c measurement at screening. That means a participant with an HbA1c of 6.6% - technically above the diagnostic threshold - could have been enrolled if they had not yet received a formal diagnosis. From STEP 1 onward, the main obesity trials adopted measured HbA1c screening and excluded anyone at or above 6.5%, making the enrolled populations metabolically cleaner in a reproducible way. TRIUMPH-1 (retatrutide, reported May 2026) reverted to the SCALE-era approach: T2D excluded by diagnosis, no HbA1c cap at screening. This design choice makes TRIUMPH-1's 28.3% headline harder to compare directly to STEP or SURMOUNT results than those trials are to each other.

Whether metabolically healthier patients respond better or worse to GLP-1 drugs is a separate question - but the populations are different, and the difference is systematic.

2. Trial duration increased by 16-29%

SCALE ran for 56 weeks. STEP and SURMOUNT ran for 68-72 weeks. The GLP-1 weight-loss curve does not plateau quickly - participants continue losing weight through week 60+ on semaglutide and tirzepatide. Running a SCALE-era trial to the longer STEP endpoint would mechanically produce a higher headline number even without changing the drug. The 16-week gap between SCALE and STEP 1 is equivalent to roughly one-quarter of the entire SCALE treatment period.

3. Comorbidity definitions evolved

SCALE accepted hypertension or dyslipidaemia as the qualifying comorbidity for BMI ≥27 enrolment. STEP 1 used a broader list of weight-related conditions. SURMOUNT-1 specified "obesity-related complications" including obstructive sleep apnoea, cardiovascular disease, and hypertension, among others. The specific comorbidities affect baseline metabolic burden, and therefore baseline responsiveness to appetite suppression.

What This Means in Practice

None of this is to say that the drugs have not improved - they clearly have. Tirzepatide (Zepbound, Mounjaro) dual GLP-1 and GIP mechanism genuinely produces more weight loss than GLP-1 alone, and the higher-dose semaglutide (Wegovy) in STEP UP has pushed the molecule meaningfully beyond its 2.4 mg ceiling. The pharmacology is real.

The problem is that the magnitude of that improvement cannot be read cleanly off the raw trial numbers. The liraglutide-to-semaglutide step, for example, conflates drug mechanism, dose, trial duration, and HbA1c screening protocol all at once. A naive reading suggests semaglutide is roughly twice as effective as liraglutide. The true pharmacological uplift is narrower - and becomes visible only in a head-to-head trial. SURMOUNT-5, which directly compared tirzepatide and semaglutide in matched patients under identical conditions, showed a ~20% edge for tirzepatide (20.2% vs 13.7%). That is the only number in this space you can trust without a footnote.

Head-to-Head Trials: The Only Clean Comparison

The table below lists all known interventional head-to-head trials comparing two or more of these molecules, sourced directly from the ClinicalTrials.gov API. Results where shown are the posted primary outcome values from the ClinicalTrials.gov results section - not press releases. Studies with no posted results are marked as pending.

Obesity (weight loss as primary endpoint)

Type 2 diabetes (HbA1c primary, weight loss secondary)

These trials enrolled people with type 2 diabetes and used glycaemic control as the primary endpoint. The T2D doses of these drugs carry different brand names: semaglutide 0.5-1 mg is Ozempic, liraglutide 1.2-1.8 mg is Victoza, and tirzepatide in the T2D indication is sold as Mounjaro. The trials are listed here because weight loss was a pre-specified secondary outcome and reported comparatively.

When head-to-head data disappointed

Head-to-head trials do not always confirm what cross-trial comparisons implied. Two examples stand out from the posted CTgov results above.

In STEP 8 (NCT04074161), liraglutide 3.0 mg (Saxenda) achieved −6.4% body weight at 68 weeks. The SCALE Obesity pivotal trial, which ran 56 weeks in a different population without a measured HbA1c screen, showed around −7.9% for the same dose. Placed head-to-head against semaglutide (Wegovy) under identical conditions and a stricter metabolic screen, liraglutide performed noticeably below its own pivotal-trial result, while the gap between the two drugs widened to nearly 10 percentage points. Cross-trial estimates had implied a gap closer to 6-7 points.

In SURMOUNT-5 (NCT05822830), semaglutide 2.4 mg (Wegovy) achieved −13.7% body weight, compared to −14.9% in STEP 1 (NCT03548935) - the same drug, same dose, in a broadly similar population, run by the same sponsor. Both results are from the same intent-to-treat analysis. The −1.2 percentage point gap between the two semaglutide arms may seem small, but it underscores a basic point: the same drug performs differently across different trial contexts. When that drug is your benchmark for evaluating a competitor, that variability matters.

The sharpest example to date is REDEFINE 4 (NCT06131437), which reported in February 2026. CagriSema - a fixed-dose combination of cagrilintide and semaglutide (Wegovy) - had looked ahead of tirzepatide (Zepbound) based on cross-trial comparison: REDEFINE 1 showed approximately 22% weight loss for CagriSema, while SURMOUNT-1 showed approximately 21% for tirzepatide. The naive cross-trial read implied CagriSema was modestly superior. In the open-label head-to-head over 84 weeks, tirzepatide achieved −25.5% versus CagriSema's −23.0%, and CagriSema failed to meet the pre-specified non-inferiority threshold. The drug that looked ahead on cross-trial data lost the direct comparison by 2.5 percentage points.

Revenue Tells a Different Story

If the trial data suggests a smooth march of progress - 8% to 15% to 22% - the revenue data is more complicated. The chart below shows quarterly worldwide sales for the three approved obesity drugs from company earnings reports.

Three things jump out. First, Saxenda grew steadily but never broke out - peak quarterly sales were around $470M, reached eight years after approval. Second, Wegovy's 2022 supply dip is visible: Novo had to restrict new prescriptions mid-launch, a commercial stumble that cost it first-mover advantage by the time Zepbound arrived. Third, Zepbound reached $517M in its second full quarter on market - a faster commercial ramp than Wegovy achieved at any comparable stage, in part because it launched into a market Wegovy had already primed.

The trial headlines said tirzepatide was better than semaglutide. The revenue data confirms the market believed it. But how much of tirzepatide's commercial advantage is pharmacology, and how much is the timing of launching second into an established category with fewer supply constraints? That question cannot be answered from the headline percentages.

The Investment Angle

Headline numbers move stock prices; protocol details determine whether the move is deserved

When a new weight-loss trial result is published, the market often reacts within minutes to the headline percentage. That number then gets compared to prior trials and a verdict is delivered: better, worse, or competitive. But as this post has shown, those comparisons are built on shaky foundations. Before drawing conclusions, four questions are worth asking:

• How long was the trial? A 48-week result cannot be compared directly to a 72-week result. These drugs keep working past week 48, so longer trials mechanically produce higher numbers - even if the drug is identical.
• How were patients screened for diabetes? SCALE used a diagnosis-based exclusion. STEP and SURMOUNT measured HbA1c at screening and excluded anyone at or above 6.5%. That difference in screening rigor changes who gets enrolled and potentially how they respond.
• Is there a head-to-head trial, or is this a cross-trial comparison? Pipeline drugs from smaller companies are often benchmarked against Wegovy or Zepbound results from different protocols. Without matching the trial design, the comparison is directional at best. The only number in this space that needs no asterisk is from a head-to-head study like SURMOUNT-5.
• What is the dropout rate? "On-treatment" figures (OT in the table above) only count patients who completed the study. "Intent-to-treat" figures (IT) include everyone who was randomised, including dropouts. The gap between the two reflects tolerability - and tolerability is what drives real-world adherence, repeat prescriptions, and ultimately revenue.

Why this matters now

Several next-generation candidates have now reported or are close to reporting. CagriSema lost the direct head-to-head in REDEFINE 4 despite looking ahead on cross-trial data. Retatrutide reported its TRIUMPH-1 pivotal result on 21 May 2026: 28.3% weight loss at 80 weeks using the efficacy estimand - but using the older diagnosis-based diabetes exclusion with no HbA1c cap at screening, the same design as SCALE in 2015. That single protocol detail makes TRIUMPH-1 the first major obesity pivotal since SCALE to use that approach, and it is a real confound when comparing its headline against the HbA1c-capped STEP and SURMOUNT results. Amycretin and the next generation of orals still have pivotal results outstanding. The pattern - large headline number, a design detail that limits comparability - is likely to repeat.

The most durable takeaway from a decade of these trials is straightforward: trust head-to-head data, treat cross-trial comparisons as directional, and always ask who was in the room - and for how long - before deciding what a headline number actually means.