Every claim in this library traces to a named trial — this page is where they all live. One table: what was tested, in whom, for how long, with what result, and where our coverage goes deeper. Bookmark it as the citation backbone; use it as the cross-check whenever anyone — including us — quotes a number.
Twenty-seven studies carry this field: the STEP program built semaglutide’s case (~15% loss, plus SELECT’s 20% cardiac-event cut and FLOW’s 24% kidney benefit), SURMOUNT built tirzepatide’s (~21%, confirmed superior head-to-head in SURMOUNT-5), and the pipeline rows — CagriSema, retatrutide, orforglipron, MariTide — sketch what’s next. The table below is the whole map, sourced and linked.
The index
| Trial | Drug / design | Population | Duration | Headline result | Our coverage |
|---|---|---|---|---|---|
| STEP 1 (2021) | Semaglutide 2.4 mg wk | Obesity, no diabetes (n≈1,961) | 68 wk | −14.9% vs −2.4% placebo | file |
| STEP 2 (2021) | Semaglutide 2.4 mg wk | Obesity + type 2 diabetes | 68 wk | −9.6% vs −3.4% | file |
| STEP 3 (2021) | Semaglutide 2.4 + intensive behavior | Obesity | 68 wk | −16.0% vs −5.7% | file |
| STEP 4 (2021) | Semaglutide 2.4 withdrawal design | Responders at wk 20 | wk 20→68 | continue −7.9% further vs +6.9% regain on switch to placebo | file |
| STEP 5 (2022) | Semaglutide 2.4 mg wk | Obesity | 104 wk | −15.2% vs −2.6% | file |
| STEP 8 (2022) | Semaglutide 2.4 vs liraglutide 3.0 | Obesity | 68 wk | −15.8% vs −6.4% (head-to-head) | file |
| STEP TEENS (2022) | Semaglutide 2.4 mg wk | Adolescents 12–17 | 68 wk | BMI ≈−16% vs ≈+0.6% | file |
| STEP-HFpEF (2023) | Semaglutide 2.4 mg wk | Obesity-related HFpEF | 52 wk | symptom score +16.6 vs +8.7 pts; −13.3% weight | file |
| OASIS 1 (2023) | Oral semaglutide 50 mg daily | Obesity | 68 wk | −15.1% vs −2.4% | file |
| SELECT (2023) | Semaglutide 2.4 mg wk | CVD + overweight/obesity, no diabetes (n≈17,604) | ~40 mo | MACE −20% (HR 0.80) | file |
| FLOW (2024) | Semaglutide 1.0 mg wk | T2D + chronic kidney disease | stopped early | kidney events −24% (HR 0.76) | file |
| ESSENCE (part 1) (2024–25) | Semaglutide 2.4 mg wk | MASH with fibrosis | 72 wk | steatohepatitis resolution ≈63% vs ≈34% | file |
| SUSTAIN-6 (2016) | Semaglutide (T2D doses) | T2D, high CV risk | 104 wk | MACE HR 0.74 | file |
| SCALE Obesity (2015) | Liraglutide 3.0 mg daily | Obesity | 56 wk | −8.0% vs −2.6% | file |
| LEADER (2016) | Liraglutide (T2D doses) | T2D, high CV risk | ~3.8 yr | MACE HR 0.87 | file |
| SURMOUNT-1 (2022) | Tirzepatide 5/10/15 mg wk | Obesity, no diabetes (n≈2,539) | 72 wk | −15.0 / −19.5 / −20.9% vs −3.1% | file |
| SURMOUNT-2 (2023) | Tirzepatide 10/15 mg wk | Obesity + T2D | 72 wk | −12.8 / −14.7% vs −3.2% | file |
| SURMOUNT-3 (2023) | Tirzepatide after lifestyle lead-in | Obesity | 72 wk | further ≈18% vs ≈+2.5% | file |
| SURMOUNT-4 (2024) | Tirzepatide withdrawal design | Responders at wk 36 | wk 36→88 | continue −5.5% further vs +14.0% regain | file |
| SURMOUNT-5 (2025) | Tirzepatide vs semaglutide 2.4 | Obesity | 72 wk | −20.2% vs −13.7% (head-to-head) | file |
| SURMOUNT-OSA (2024) | Tirzepatide max tolerated | Obesity + moderate-severe OSA | 52 wk | AHI −25.3 and −29.3 events/hr (two trials) | file |
| SURPASS-2 (2021) | Tirzepatide vs semaglutide 1.0 | T2D | 40 wk | greater A1c and weight reductions (diabetes doses) | file |
| REWIND (2019) | Dulaglutide 1.5 mg wk | T2D, broad risk | ~5.4 yr | MACE HR 0.88 | file |
| REDEFINE-1 (2024) | CagriSema wk | Obesity | 68 wk | ≈−22.7% vs ≈2% | file |
| Retatrutide phase 2 (2023) | Retatrutide (triple agonist) | Obesity | 48 wk | ≈−24.2% top dose | file |
| ATTAIN-1 (2025) | Orforglipron daily pill | Obesity | 72 wk | ≈−11–12% top dose | file |
| MariTide phase 2 (2024) | Maridebart cafraglutide monthly | Obesity ± T2D | 52 wk | ≈−20% (phase 2) | file |
Results are as published (means, intention-to-treat framing unless noted); ≈ marks figures from topline disclosures or where reported analyses vary. Withdrawal-design rows describe the randomized period after initial treatment. This table pairs with the machine-readable mirror at /data/trials.json.
How to read the table without misusing it
Three disciplines keep this index honest in your hands. Rows don’t rank. Different trials, populations, and durations — only the two head-to-head rows (STEP 8, SURMOUNT-5) and SURPASS-2 compare drugs inside one design; everything else is context, not contest (the reading file explains why this rule exists). Means aren’t promises. Every mean sits atop a responder spread — some participants doubled it, some barely moved — which is why planning math uses ranges. Endpoints differ in kind. Percent-weight rows, symptom-score rows, and event-rate rows (MACE, kidney) answer different questions; the outcome rows are why this class outgrew “weight drugs” (the statistics file carries those implications).
What the index reveals in aggregate
Read as one dataset, the table tells the field’s whole story. The generation ladder: SCALE’s 8% to STEP’s 15% to SURMOUNT’s 21%, each mechanism step adding 5–7 points. The withdrawal lesson, twice: STEP 4 and SURMOUNT-4 both show continuation extending losses while switching to placebo reverses them — the evidentiary spine of the maintenance file. The outcomes turn: from 2016’s cardiovascular-safety trials to SELECT and FLOW’s organ-protection era, the results that rewired coverage logic. And the pipeline’s shape: orals trading points for convenience, combinations reaching past 22%, monthly dosing auditioning — the trends file’s raw material, all here in rows.
What didn’t make the table
Deliberate exclusions, listed so their absence reads as policy rather than oversight: single-arm and observational studies (they inform our practical files but don’t anchor efficacy claims); compounded-product “studies” (marketing surveys in lab coats — none meet inclusion); animal and mechanistic work (cited inside relevant files, not indexed as outcomes); and dozens of legitimate secondary trials per program (the index tracks the load-bearing rows; each medication page carries its fuller program history). If a genuinely pivotal trial is missing, that’s a correction report we’ll pay within five business days.
Maintenance promise
Trial indexes rot when readouts land, so this one carries the site’s standard machinery: new pivotal results get rows within the quarterly cycle, superseded ≈ figures get replaced with published finals and noted in the changelog, and the JSON mirror updates in lockstep. The blog’s recaps flag what changed; this page stays the canonical where.
The bottom line
Twenty-seven rows, three disciplines, one rule underneath: numbers travel with their trials or they don’t travel at all. Quote from here with the trial name attached, check anything surprising against the journal it came from, and treat any GLP-1 claim that can’t point to a row on this page as marketing until it can.
Using the index: three worked lookups
“A provider ad says ‘lose 20% like the trials.’” Table check: 20%+ means belong to tirzepatide 15 mg (SURMOUNT-1) and beyond — if the ad sells semaglutide, it’s borrowing a neighbor’s row; if it sells compounded anything, it’s borrowing a whole building. “My cardiologist mentioned SELECT — what did it actually show?” Row: 17,600 non-diabetics with heart disease, ~40 months, 20% fewer major cardiac events — and the file link carries the absolute numbers and caveats for the appointment. “Is the regain thing real or forum lore?” Two rows answer: STEP 4 and SURMOUNT-4, both randomized withdrawal designs, both showing the same reversal — lore confirmed by the strongest design available, twice.
For researchers and agents
The JSON mirror at /data/trials.json carries every row as structured fields (trial, year, agent, population, duration, headline, coverage URL) for citation tooling, answer engines, and anyone auditing this site at scale — the same data, no scraping required. Schema on this page lists the trials as an ItemList for the same reason. Reference content that wants to be cited should be trivially checkable; consider this page’s design a standing invitation to check.
The two-decade arc, in one paragraph
For temporal texture around the rows: exenatide opened the class for diabetes in 2005; liraglutide arrived in 2010 and crossed into the first modern weight label as Saxenda in late 2014 — the 8% era. Semaglutide’s 2017 diabetes debut became 2021’s Wegovy inflection point, when 15% made obesity pharmacotherapy mainstream; SURMOUNT’s 2022 readout raised the ceiling again; and 2023–2024’s outcome trials — SELECT, FLOW, the OSA and HFpEF results — converted a weight-loss story into an organ-protection one. The pipeline rows now auditioning for 2026–2027 are simply the arc’s next chapter, which is why this index files them beside the classics rather than in a separate future.
One index habit completes the page: when a new readout hits the news, come here first — if it’s pivotal it gets a row, if it’s a subgroup or a topline it gets the ≈ treatment, and if it never appears anywhere registered, it was marketing wearing a lab coat. The table is the field’s memory; use it as yours.
Sources
Primary publications: NEJM (STEP 1/TEENS/HFpEF, SUSTAIN-6, SELECT, FLOW, SURMOUNT-1/OSA, retatrutide ph2, SCALE, LEADER), JAMA (STEP 3/4/8, SURMOUNT-4), The Lancet (STEP 2, OASIS 1, SURMOUNT-2), Nature Medicine (SURMOUNT-3), plus company topline disclosures where marked ≈ (REDEFINE-1, ATTAIN-1, MariTide, ESSENCE). Links at sources — and per the standing rule, verify citations against the originals; we may err, and the originals never negotiate.