Beyond diabetes and obesity: where GLP-1 drugs are being tested in 2026

GLP-1 drugs were developed for type 2 diabetes and became famous for weight loss. The clinical trial registry shows where they are heading next. Of the 1,543 trials on ClinicalTrials.gov that test a GLP-1 drug, 340 study a disease other than diabetes or obesity (not counting phase 1 pharmacology studies), and their share is rising. Of the trials due to start in 2026, 76 are in these newer areas and 90 in diabetes and obesity.

Every number here was produced by ClinLit AI from the registry and PubMed, and every claim about a protocol comes from the protocol itself. Each section has a Reproduce this with ClinLit AI box with the question we asked, an excerpt of the answer and the queries behind it, so you can run the same research or adapt it to your own drug or disease. The full walkthrough is in How we researched this post with ClinLit AI.

Key findings

• The shift is recent. From 2010 to 2019, about one in four new GLP-1 trials (excluding phase 1 pharmacology studies) was outside diabetes and obesity. From 2024 to 2026 it is more than one in three.
• Four areas already have phase 3 results and new approved uses: cardiovascular risk (March 2024), obstructive sleep apnoea (December 2024), chronic kidney disease in type 2 diabetes (January 2025) and MASH (August 2025).
• Addiction is the fastest-growing area: 51 trials, 36 of them started since 2024, almost all run by universities and public funders rather than drug companies.
• Cancer trials are new and academic: 25 trials, 21 started since 2024, many testing weight loss with a GLP-1 drug alongside fertility-sparing treatment for early endometrial cancer.
• Not everything works. Two large phase 3 programmes in neurodegenerative disease, in Parkinson’s disease and in Alzheimer’s disease, found no slowing of progression.
• The protocols show the field learning how to run these trials. New indications need new endpoints, and several flagship trials changed their primary endpoints partway through.

The trend

Where the trials are

Each trial is counted once, in the first area it matches (the definition is in the box below). “Open” means recruiting, about to recruit, or still running, as of the 1 October 2026 snapshot.

For comparison: type 2 diabetes accounts for 540 trials (118 started since 2024), obesity for 234 (141) and type 1 diabetes for 63 (20). Another 335 are phase 1 pharmacology studies, mostly in healthy volunteers, and 31 fit none of the areas.

Who runs the trials depends on the area. Sleep apnoea is almost entirely industry: 14 of the 15 trials started since 2020 are led by a company, and 15 of the 18 are phase 3. Addiction and cancer are the opposite: only 4 of 44 addiction trials, and none of the 23 cancer trials, started since 2020 are led by a company. Most are small phase 2 studies run by universities, hospitals and public funders.

The drugs have changed too. Of the 167 trials in these newer areas started since 2024, 64 include semaglutide, 50 tirzepatide and 48 a newer agent, such as mazdutide, HRS9531, survodutide, orforglipron, maridebart cafraglutide, pemvidutide or retatrutide. The older drugs that opened these areas, exenatide and liraglutide, appear in one each.

WITH glp AS (
  SELECT DISTINCT i.nct_id
  FROM aact.interventions i
  JOIN aact.studies s ON s.nct_id = i.nct_id
  WHERE s.study_type = 'INTERVENTIONAL'
    AND i.intervention_type IN ('DRUG','BIOLOGICAL','COMBINATION_PRODUCT','OTHER')
    AND (
      i.name ~* 'semaglutide|liraglutide|tirzepatide|dulaglutide|exenatide|lixisenatide|albiglutide|efpeglenatide|beinaglutide|loxenatide|taspoglutide|orforglipron|retatrutide|survodutide|cotadutide|mazdutide|amycretin|cagrisema|maridebart|maritide|pemvidutide|efinopegdutide|ecnoglutide|danuglipron|lotiglipron|dapiglutide|efocipegtrutide|bofanglutide|supaglutide|glutazumab|aleniglipron|pegapamodutide|ozempic|wegovy|rybelsus|victoza|saxenda|trulicity|byetta|bydureon|mounjaro|zepbound|lyxumia|adlyxin|soliqua|xultophy|ly3298176|ly3437943|ly3502970|bi 456906|ibi362|amg 133|medi0382|itca 650|hrs9531|hrs-7535|vk2735|ct-388|ct-996|pf-06882961|xw003|gzr18|tg103|nnc0519'
      OR (i.name ~* '(glp-?1|glucagon-like peptide[- ]?1)' AND i.name ~* '(agonist|analog|\yra\y)' AND i.name !~* 'antagon')
    )
),
cond AS (
  SELECT nct_id, string_agg(lower(name), ' ') AS names FROM aact.conditions GROUP BY nct_id
),
base AS (
  SELECT g.nct_id, s.phase, s.overall_status, s.start_date,
    lower(coalesce(s.brief_title,'') || ' ' || coalesce(s.official_title,'') || ' ' || coalesce(c.names,'')) AS t
  FROM glp g
  JOIN aact.studies s ON s.nct_id = g.nct_id
  LEFT JOIN cond c ON c.nct_id = g.nct_id
),
area AS (
  SELECT *, CASE
    WHEN phase IN ('PHASE1','EARLY_PHASE1') AND t ~ 'healthy|pharmacokinetic|bioequivalen|drug interaction|renal impairment|hepatic impairment|formulation' THEN '01 phase 1 pharmacology'
    WHEN t ~ 'nash|\ymash\y|steatohepatitis|fatty liver|steatotic|nafld|masld|cirrhosis' THEN '02 liver'
    WHEN t ~ 'alcohol|opioid|substance use|smoking|nicotine|tobacco|cocaine|methamphetamine|cannabis' THEN '03 addiction'
    WHEN t ~ 'alzheimer|parkinson|dementia|cognitive impairment' THEN '04 neurodegenerative'
    WHEN t ~ 'kidney|renal|\yckd\y|nephropathy|dialysis|albuminuria' THEN '05 kidney'
    WHEN t ~ 'heart failure|cardiovascular|myocardial|coronary|atherosclerosis|stroke|peripheral arterial|atrial fibrillation' THEN '06 heart and vessels'
    WHEN t ~ 'apnoea|apnea' THEN '07 sleep apnoea'
    WHEN t ~ 'osteoarthrit' THEN '08 osteoarthritis'
    WHEN t ~ 'pcos|polycystic ovar|infertil' THEN '09 pcos infertility'
    WHEN t ~ 'schizophreni|psychosis|bipolar|antipsychotic|depressi|binge eating' THEN '10 mental health'
    WHEN t ~ 'cancer|tumour|tumor|carcinoma|neoplas|endometrial hyperplasia' THEN '11 cancer'
    WHEN t ~ 'psoriasis|hidradenitis|crohn|colitis|rheumatoid|asthma|copd' THEN '12 inflammatory'
    WHEN t ~ 'type 1 diabet|type i diabet|\yt1d\y|t1dm|diabetes mellitus type 1|juvenile-onset diabet|insulin-dependent diabet' THEN '13 type 1 diabetes'
    WHEN t ~ 'diabet|glyc|hba1c|glucose' THEN '14 type 2 diabetes'
    WHEN t ~ 'obes|overweight|weight|\ybmi\y' THEN '15 obesity'
    ELSE '16 other' END AS area
  FROM base
)
SELECT area,
  count(*) AS total,
  count(*) FILTER (WHERE start_date >= DATE '2024-01-01') AS started_2024_or_later,
  count(*) FILTER (WHERE overall_status IN ('RECRUITING','NOT_YET_RECRUITING','ACTIVE_NOT_RECRUITING','ENROLLING_BY_INVITATION')) AS open_now,
  count(*) FILTER (WHERE phase IN ('PHASE3','PHASE2/PHASE3')) AS phase3_or_23
FROM area
GROUP BY area
ORDER BY area;

Area by area

Heart and blood vessels

The largest area, and the one that changed GLP-1 prescribing first. SELECT (NCT03574597) randomised 17,604 people with overweight or obesity and established cardiovascular disease, without diabetes, to semaglutide or placebo. Major cardiovascular events fell by 20% (hazard ratio 0.80) (PMID 37952131), and the FDA approved Wegovy to reduce that risk in March 2024.

Heart failure with preserved ejection fraction (HFpEF) followed. STEP-HFpEF (NCT04788511) improved symptoms and physical limitation with semaglutide (PMID 37622681), and SUMMIT (NCT04847557) found tirzepatide lowered the risk of cardiovascular death or worsening heart failure (hazard ratio 0.62) (PMID 39555826).

The newer trials reach further: acute ischaemic stroke treated with thrombectomy, intracerebral haemorrhage, atrial fibrillation after ablation, peripheral artery disease, and cardiovascular outcome trials for the next generation of drugs, such as orforglipron and maridebart cafraglutide.

Liver disease (MASH)

Most liver trials target metabolic dysfunction-associated steatohepatitis (MASH, formerly NASH), and most are industry-run (20 of the 34 started since 2020), because liver biopsy endpoints are expensive. ESSENCE (NCT04822181) is the reference trial: at 72 weeks, MASH resolved without worsening fibrosis in 62.9% of patients on semaglutide versus 34.3% on placebo (PMID 40305708). That led to accelerated approval of Wegovy for noncirrhotic MASH with moderate to advanced fibrosis in August 2025. Its second part runs to 240 weeks, with cirrhosis-free survival as the endpoint, to show whether the histology turns into fewer liver events.

Phase 2 trials of dual agonists set up the next wave. Tirzepatide (SYNERGY-NASH, NCT04166773, PMID 38856224) and survodutide (NCT04771273, PMID 38847460) both reported large improvements in MASH. Phase 3 programmes are now running for survodutide (LIVERAGE, NCT06632444, 1,800 participants, plus a separate trial in cirrhosis) and pemvidutide (PERFORMA, NCT07795164).

Kidney disease

FLOW (NCT03819153) was the first dedicated kidney outcome trial of a GLP-1 drug. In 3,533 people with type 2 diabetes and chronic kidney disease, semaglutide lowered the risk of the primary kidney and cardiovascular composite by 24% (hazard ratio 0.76), and the trial stopped early for efficacy (PMID 38785209). Ozempic was approved for this use in January 2025. The newer trials test GLP-1 drugs with finerenone and SGLT2 inhibitors, in type 1 diabetes, in polycystic kidney disease, and after transplantation.

Obstructive sleep apnoea

A small area that went straight to phase 3. SURMOUNT-OSA (NCT05412004) cut the apnoea-hypopnoea index (AHI) by 20 to 24 more events per hour than placebo with tirzepatide (PMID 38912654). Zepbound became the first drug approved for the condition in December 2024. Eleven of the 12 sleep apnoea trials started since 2024 test newer agents, among them retatrutide, orforglipron, maridebart cafraglutide and several drugs developed in China.

Knee osteoarthritis

STEP 9 (NCT05064735) showed that semaglutide reduced knee pain (WOMAC pain score) as well as weight in people with obesity and knee osteoarthritis (PMID 39476339). Lilly is now testing orforglipron for osteoarthritis pain (ATTAIN-OA PAIN, NCT07153471).

Addiction

The fastest-growing area, and the least settled. The 51 trials cover alcohol, opioids, cocaine, methamphetamine, cannabis and nicotine, and most are small, investigator-led phase 2 studies. Alcohol leads: ClinLit AI found 22 trials of GLP-1 drugs in alcohol use disorder.

The early evidence is mixed. A Danish trial of exenatide in alcohol use disorder (NCT03232112) missed its primary endpoint, heavy drinking days, though exploratory analyses suggested an effect in people with obesity (PMID 36066977). A US trial of low-dose semaglutide (NCT05520775) reduced the amount of alcohol people drank in a laboratory session and their drinks per drinking day (PMID 39937469). In July 2026, Altimmune reported that pemvidutide reduced heavy drinking days in its phase 2 RECLAIM trial (NCT06987513; company press release, not yet peer reviewed). Lilly’s phase 2 trial of mazdutide in alcohol use disorder (NCT06817356) finished in March 2026, and we found no published results yet.

Larger trials are starting, including a 622-person US Veterans Affairs trial (CRAVE, NCT07218354) and a 310-person trial of tirzepatide alongside buprenorphine for opioid use disorder (NCT06651177).

Cancer

Twenty-one of the 25 cancer trials started in 2024 or later, and none of those started since 2020 is run by a company. The largest group, 12 trials, tests whether weight loss with a GLP-1 drug helps fertility-sparing treatment of atypical endometrial hyperplasia and early endometrial cancer, mostly with tirzepatide and a levonorgestrel intrauterine device, and endpoints such as complete pathological response (for example WE-FiERCE, NCT06073184). Others test tirzepatide in early breast cancer (TRIM-EBC, NCT06517212, with clearance of circulating tumour DNA and distant disease-free survival as primary endpoints) and semaglutide before prostate surgery.

The literature has not caught up. Of the 148 PubMed articles on GLP-1 drugs whose main topics include cancer, 92 are about pancreatic or thyroid tumours, largely questions of safety, and 6 are about endometrial cancer.

Neurodegenerative disease

The cautionary area. Exenatide-PD3 (NCT04232969) found no difference from placebo in Parkinson’s motor scores after 96 weeks (PMID 39919773). EVOKE and EVOKE+ (NCT04777396, NCT04777409), which together randomised about 3,800 people with early Alzheimer’s disease, did not slow decline on the Clinical Dementia Rating, although Alzheimer’s biomarkers improved (Novo Nordisk, 24 November 2025). Only five trials in this area have started since 2024.

Smaller areas

• PCOS and fertility (17 trials): ovulation and fertility outcomes, such as RESTORE (NCT05819853), whose primary endpoint is ovulation frequency.
• Inflammatory disease (12): psoriasis, psoriatic arthritis and Crohn’s disease, including Lilly trials that pair tirzepatide with its own biologics, ixekizumab and mirikizumab (TOGETHER-PsO, NCT06588283, whose primary endpoint combines complete skin clearance with at least 10% weight loss).
• Mental health (11): weight gain on antipsychotics, cognition in depression, and binge eating disorder.
• One or two trials each in idiopathic intracranial hypertension, long COVID, lymphoedema, short bowel syndrome, trichotillomania and biological ageing.

What the protocols show about study design

Many sponsors post their protocols and statistical analysis plans (SAPs) on ClinicalTrials.gov, and ClinLit AI can read them. We asked it about the flagship trials in these areas and checked every passage it quoted against the PDF. Together, they show what it takes to move a drug into a new indication.

Each new indication needs its own endpoint

Two patterns stand out. First, weight often sits beside the disease endpoint: STEP-HFpEF and STEP 9 made body weight a primary endpoint alongside symptoms or pain, and TOGETHER-PsO combined skin clearance with 10% weight loss in a single responder endpoint. Second, outcome trials are event-driven. FLOW’s protocol required at least 854 primary endpoint events, sized on event rates in the matching subgroup of the earlier LEADER trial, and planned interim analyses that allowed an early stop for efficacy, which is what happened (FLOW protocol, sections 5.1 and 10.1, PDF pp. 20 and 47).

Endpoints changed partway through

New indications have no settled regulatory endpoint, and the amendment histories show it.

• SUMMIT (tirzepatide in HFpEF) rewrote its primary endpoints in protocol amendment (c), approved 14 February 2024. A hierarchical composite assessed by win ratio, and the 6-minute walk distance, left the primary analysis. Two primaries replaced them: the KCCQ clinical summary score, and cardiovascular death or heart failure events. The amendment also broadened “heart failure event” to include worsening treated with oral diuretics as an outpatient. The protocol gives two reasons: the original endpoints “may not be able to capture the full potential benefit offered by tirzepatide”, and the win-ratio components “present significant complexity in analytics and interpretation” (SUMMIT protocol, PDF pp. 4–5 and 12).
• SURMOUNT-OSA moved in two steps. Amendment (b), in September 2022, added the absolute change in AHI as a key secondary endpoint “to align with recommendations from health authority”. Amendment (c), approved 2 June 2023, made it the primary endpoint in place of the percent change, “in response to regulatory recommendation”, and the two swapped places (SURMOUNT-OSA master protocol, PDF pp. 4 and 113).

Who gets in: obesity is usually an entry criterion

Most of these trials enrol only people with obesity:

• STEP-HFpEF required a BMI of at least 30 and excluded people with diabetes, who went into a separate trial. Its NT-proBNP thresholds are lower for people with a BMI of 35 or more (protocol, PDF pp. 25–26).
• SUMMIT also required a BMI of at least 30, an age of 40 or more, an ejection fraction of at least 50%, and either reduced kidney function or a recent heart failure decompensation (protocol, PDF pp. 52–53).
• STEP 9 required a BMI of at least 30, knee osteoarthritis of Kellgren-Lawrence grade 2 or 3, and a WOMAC pain score of at least 40 (protocol).
• SURMOUNT-OSA required a BMI of at least 30 and an AHI of at least 15 on polysomnography, and ran as two trials, for people using positive airway pressure and for people not using it (protocol, PDF p. 32).
• SYNERGY-NASH allowed a BMI from 27 to 50, with biopsy-confirmed MASH at fibrosis stage 2 or 3 (protocol, PDF pp. 25–26).
• Even the addiction trials follow suit: SEMALCO enrols people with alcohol use disorder and a BMI of at least 30 (protocol, PDF p. 3), and RECLAIM required a BMI over 25.

FLOW is the exception. It has no BMI criterion: it selected participants on kidney measures (eGFR and urine albumin-to-creatinine ratio, in two defined bands) and capped the share with an eGFR of 60 or more at 20% to keep the event rate high (FLOW protocol, sections 5.1 and 6.1).

These criteria are sensible: obesity drives many of these diseases, and the drugs are approved for weight management. But they shape what the results can show. When every participant has obesity and loses substantial weight, it is hard to say how much of a benefit is a direct effect of the drug and how much follows from the weight loss. Several trials now publish analyses by weight change for this reason.

Screening is expensive, so protocols try to make it cheaper

Biopsy-based liver trials lose many candidates at screening. In SYNERGY-NASH, early blinded biopsy data from the trial itself showed that a FAST score, calculated from FibroScan and AST, would cut the number of biopsies that failed to qualify. An amendment made a FAST score above 0.35 a condition for the screening biopsy and raised the minimum AST from 20 to 23 U/L (protocol, PDF pp. 3, 25 and 48). SUMMIT qualified participants with repeated 6-minute walk tests, which had to fall between 100 and 425 metres and agree closely with each other, to keep the baseline stable (protocol, PDF p. 52).

Sizing a trial with little prior data

The addiction protocols are candid about how little evidence they started from.

• The US semaglutide trial in alcohol use disorder had “no a priori effect size estimate”. It was powered to detect a medium effect with 36 participants completing treatment, allowed for 25% dropout and randomised 48. It enrolled only people not seeking treatment for their drinking (protocol and SAP, PDF pp. 6 and 42).
• SEMALCO borrowed its effect size from a trial of psilocybin-assisted therapy, because no semaglutide data existed. That gave 64 participants, which it raised to 108 assuming 40% dropout. The team’s earlier exenatide trial had lost 54.3% of participants, and its brain-imaging substudy was enlarged to 70 because only 22 of 62 people had been rescanned last time (protocol, PDF pp. 23–24).
• SEMALCO also shows the practical side of blinding an injectable: because the active and placebo pens look different and the active pen clicks, participants are blindfolded during injections and given music to listen to (PDF p. 17).

What the literature says

PubMed shows the same shift, a few years behind the registry. ClinLit AI counted articles about a GLP-1 drug (as a listed substance or main topic) whose main topics include each area. An article can count in more than one area.

The overall volume has taken off: 639 articles on GLP-1 drugs from 2023, 1,145 from 2024 and 1,884 from 2025.

How we counted

• Trials: interventional trials on ClinicalTrials.gov in the AACT snapshot of 1 October 2026, which leaves out observational studies and phase 4 trials. A trial counts if any intervention’s name matches a GLP-1 receptor agonist, including dual and triple agonists such as tirzepatide, retatrutide and survodutide, by generic name, brand name or development code, or names the GLP-1 drug class. It counts even when the drug is the comparator. The full definition is in the first box above.
• Areas: each trial goes into one area, the first whose keywords appear in its titles or condition names, in a fixed order, so a trial in people with type 2 diabetes and kidney disease counts as kidney disease. This is keyword matching, not manual review: some trials will be in the wrong area.
• Years: by start date. For trials not yet recruiting, that is the planned start.
• Literature: articles in the 2026 PubMed baseline (published since 2010, with an abstract), which runs to late 2025.
• Results and approvals after these data come from company and FDA announcements, linked where cited. Company topline results are not peer reviewed.
• Cross-check: we also wrote our own SQL against the same database. It found 1,555 trials rather than 1,543, and area counts within a few trials of these, because it also matched MeSH condition terms. The walkthrough explains the differences.

Every trial and article cited here can be opened from its NCT ID or PMID. Ask ClinLit AI the questions below to go further.