Semaglutide vs Tirzepatide: How to Compare Weight-Loss Trial Results

Compare STEP 1, SURMOUNT-1 and SURMOUNT-5 correctly. Understand weight-loss percentages, trial populations, dose arms, estimands and head-to-head evidence.

A headline percentage is only the beginning of a useful comparison. Before deciding what a weight-management study shows, check who participated, which treatment strategy was tested, how long participants were followed, and which analysis produced the quoted number.

SURMOUNT-5 directly randomized participants between tirzepatide and semaglutide. Original abstract Start with this comparison before placing unrelated trial percentages side by side.

The purpose of this article is to help you read those comparisons accurately. The dose amounts below identify research arms. They are not instructions for selecting or changing an individual’s medication.

First identify the question behind the trial name

Start a comparison table before collecting percentages. Give each study its own row, and keep the population and control visible. Do not combine trial names that happen to include the same medicines.

STEP 1 studied semaglutide against placebo in adults without diabetes over 68 weeks. SURMOUNT-1 studied three tirzepatide dose groups against placebo over 72 weeks, also excluding diabetes. Their numerical results came from separate randomized experiments. STEP 1, SURMOUNT-1

For your reading notes, distinguish “drug versus its trial’s placebo” from “drug versus the other drug in the same trial.” The first answers a question about that trial’s treatment comparison. To support a direct superiority claim between two medicines, look first for the trial that actually assigned participants between those alternatives.

A compact map of the major comparisons

The table preserves the study context alongside the figures. Negative values indicate average percentage change from starting body weight. They are group estimates, not forecasts for a particular person.

Sources: STEP 1 original abstract, SURMOUNT-1 original abstract, SURMOUNT-5 original abstract. The first two rows are separate placebo-controlled trials; the last is the randomized head-to-head comparison.

Do not read down the drug column as if all participants belonged to one experiment. Use this table as an index to the papers. A useful next step is to add the eligibility criteria and analysis definition to the row you plan to cite.

What the head-to-head result adds

SURMOUNT-5 was open-label, enrolled 751 participants and was funded by Eli Lilly. The 95% confidence intervals around the table’s estimates were -21.4% to -19.1% for tirzepatide and -14.9% to -12.6% for semaglutide. The comparison was statistically significant. Original abstract

Preserve “maximum tolerated” when describing its treatment arms. Do not write as if every participant necessarily received the highest listed dose. Also preserve “open-label” rather than calling the study blinded.

Our recommendation is to lead a comparison with the randomized head-to-head finding, then explain its scope. Avoid turning an average advantage in a defined trial into a guarantee that one option will be best for every patient. Individual treatment selection requires information that a headline percentage does not provide.

Percent and percentage points answer different questions

Here is an original arithmetic exercise using fictional treatments A and B. Assume their average reductions are 20% and 14% of starting weight. These are invented rounded values for teaching, not a reanalysis of a clinical trial.

The absolute difference is six percentage points: 20 minus 14. The reduction with A is approximately 43% larger relative to B’s reduction: six divided by 14, multiplied by 100. Neither description means that A removes an additional 43% of a person’s starting body weight.

For an equally fictional starting weight of 100 kilograms, a 20% reduction corresponds to 20 kilograms and a 14% reduction to 14 kilograms. The gap is six kilograms in that example. If the starting weight were 80 kilograms, those same percentages would correspond to 16 and 11.2 kilograms. The arithmetic changes with the starting value.

When you see “more weight loss,” ask whether the author means percentage points, kilograms, or a relative comparison of two reductions. Write the denominator beside the claim. A large relative percentage can be mathematically correct while still being easy to misunderstand.

Do not replace a paper’s adjusted comparison with subtraction of rounded values. STEP 1 reports an estimated treatment difference of -12.4 percentage points even though subtracting its displayed group means, -14.9 and -2.4, gives -12.5. Use the published estimate when citing the trial’s comparison. STEP 1 results

Read the analysis label beside the number

An estimand specifies the treatment effect a trial aims to estimate. The ICH E9(R1) framework distinguishes strategies for events such as treatment discontinuation. A treatment-policy strategy and a hypothetical strategy can address different questions; missing measurements still need to be handled appropriately for the chosen question. ICH guideline

In plain reading terms, ask whether the analysis aims to describe results regardless of stopping treatment, or under a defined hypothetical scenario about continued treatment. Copy the paper’s wording before shortening it. Do not assume that “on treatment,” “completed treatment,” and “hypothetical adherence” are interchangeable labels.

The STEP 1 figures above are reported for its treatment-policy estimand. The SURMOUNT-1 figures are reported for its treatment-regimen estimand. Those labels belong with the numbers. STEP 1, SURMOUNT-1

For SURMOUNT-5, this article uses the result stated in the original abstract. We did not review the complete statistical analysis plan and do not assign additional analysis details beyond the material accessed.

Why a diabetes trial may answer a different question

SURPASS-2 randomized 1,879 people with type 2 diabetes in an open-label, 40-week study. It compared tirzepatide groups with semaglutide at 1 mg, and the primary outcome concerned glycated hemoglobin, or HbA1c. That comparator and primary question differ from the obesity head-to-head trial described above. Original trial abstract

When someone cites “the semaglutide comparison,” ask which study they mean. Record the comparator amount and participant population before discussing the result. Do not silently substitute a different semaglutide arm or portray a diabetes trial as if it enrolled the same population as an obesity trial without diabetes.

This is not a reason to dismiss diabetes research. It is a reason to retain the details that make a finding interpretable. A careful comparison can discuss both studies while keeping their questions separate.

Average loss and responder rates are different summaries

Use another fictional example. Four participants lose 0%, 10%, 20% and 30% of their starting weight. Their simple average reduction is 15%, but only two of the four have reductions of at least 15%. No participant in this constructed group lost exactly the average amount.

This exercise is a reminder to request both the group summary and the outcome threshold of interest. When reading a responder rate, record the threshold, time point and denominator. Avoid changing “at least 15%” into “about 15%,” or assuming that a percentage of participants refers to everyone originally enrolled.

If a paper reports several thresholds, keep their labels attached. For a personal discussion, bring the actual trial table to a qualified clinician rather than treating the mean as a promised result. The example above is arithmetic only and does not describe the distribution in any cited trial.

Put tolerability and follow-up beside weight change

SURMOUNT-5 most commonly reported gastrointestinal events, generally mild to moderate and concentrated during escalation. Original abstract Consult the full safety record for a detailed comparison.

In a comparison worksheet, add columns for discontinuation, adverse events and observation time. Ask whether the quoted weight result includes information after participants stopped the assigned treatment. If the source available to you does not answer that question, mark it unresolved rather than selecting the most favorable interpretation.

Also record when the main measurement occurred. Do not treat a result during a study as evidence that the same change persisted indefinitely after treatment ended. Look for a separate follow-up report if maintenance after stopping is the question you want answered.

For treatment decisions, discuss expected benefits, medical history, tolerability and ongoing follow-up with the prescribing clinician. This article does not rank unregulated products or establish that a supplier’s material is equivalent to a trial medicine.

A comparison checklist you can reuse

Before sharing a weight-loss chart or citing a trial in an article, complete these fields:

1. Study identity: Full title, year and original source.
2. Participants: Diabetes status, eligibility and starting characteristics.
3. Treatment strategy: Exact comparator, dose-arm definition and accompanying intervention.
4. Duration: Time point for the quoted result.
5. Outcome: Percentage weight change, kilograms, responder rate or another measure.
6. Analysis: The estimand and handling of treatment discontinuation, when reviewed.
7. Uncertainty: Published confidence intervals and adjusted comparison.
8. Tolerability: Adverse events, withdrawals and relevant access limits.

Then label your conclusion “within-trial comparison” or “description of separate studies.” This original editorial checklist is a way to preserve context; it is not a validated treatment-selection instrument.

Frequently asked questions

Can I compare the largest percentages from different trials?

You can report them with their context, but do not present that selection as a randomized comparison between medicines. Start with the actual head-to-head study when it addresses the relevant question, and keep separate-trial results labeled as separate.

Does more average weight loss mean a treatment is best for me?

Do not make that decision from an average alone. Bring the study and your medical circumstances to the prescribing clinician. This article explains research comparisons, not individual eligibility or a dosing plan.

What if a comparison is not statistically significant?

Read the effect estimate and uncertainty before interpreting it. Our negative-study guide explains why a threshold alone is an incomplete summary. For claims involving mixtures, also see the peptide blend evidence guide.

Sources and editorial method

This is targeted research education, not a systematic review or current prescribing guide. We checked original indexed abstracts and accessible publisher result excerpts for STEP 1, SURMOUNT-1 and SURPASS-2, the complete original abstract for SURMOUNT-5, and official ICH/FDA resources on estimands. Complete trial reports and analysis plans were not appraised in full.

• STEP 1, Wilding et al., 2021
• SURMOUNT-1, Jastreboff et al., 2022
• SURMOUNT-5, Aronne et al., 2025
• SURPASS-2, Frías et al., 2021
• FDA: ICH E9(R1) estimands guidance

Confidence labels concern the reported facts or arithmetic, not a guarantee about a medicine.