Evidence review · 11 min read

GLP-1 Medicines: Benefits, Risks, and New Research

What trials establish beyond weight loss, what new infection studies suggest, and which risks matter for semaglutide and tirzepatide.

AnabioticaSeptember 4, 2026

AI-generated illustration of an unbranded capped injection pen beside a glass dish on a laboratory bench

Do GLP-1 medicines improve health beyond lowering blood sugar and body weight? For some outcomes, the answer is already yes: trials have demonstrated cardiovascular and kidney benefits in defined high-risk populations. For others, including protection against infections, the evidence is encouraging but less conclusive. The useful question is which drug, for whom, and against which outcome—not how many possible benefits can be collected under one label. (SELECT, 2023; FLOW, 2024; Han et al., 2025)

What “GLP-1” Means

GLP-1 is a hormone involved in glucose regulation and appetite. Medicines that activate its receptor can increase glucose-dependent insulin secretion, reduce appetite, and slow stomach emptying. Semaglutide is a GLP-1 receptor agonist; tirzepatide activates both GLP-1 and GIP receptors. They belong in the same conversation, but findings for one medicine, formulation, or dose do not automatically apply to the others. These are prescription treatments with specific indications. (MHRA patient guidance; Nicholls et al., 2025)

Their effects on weight can be substantial. In the 72-week, open-label SURMOUNT-5 trial of 751 adults with obesity without diabetes, average weight reduction was 20.2% with tirzepatide versus 13.7% with semaglutide. The comparison used maximum tolerated weekly doses of 10 or 15 mg and 1.7 or 2.4 mg, respectively. These are trial averages, not individual guarantees, and greater weight loss alone does not establish superiority for heart attacks, survival, or every other outcome. (Aronne et al., 2025)

Benefits Beyond the Scale

SELECT randomized 17,604 adults with overweight or obesity and established cardiovascular disease, but no diabetes, to semaglutide 2.4 mg weekly or placebo. Over about 3.3 years, cardiovascular death, nonfatal heart attack, or nonfatal stroke occurred in 6.5% versus 8.0%. The hazard was 20% lower; the observed absolute difference was 1.5 percentage points, or roughly 15 fewer events per 1,000 participants. This was prevention of further events in people already at high risk, not a trial in healthy adults. (Lincoff et al., 2023)

In FLOW, 3,533 people with type 2 diabetes and chronic kidney disease received semaglutide 1 mg weekly or placebo. Over a median 3.4 years, the primary composite occurred in 331 versus 410 participants: a hazard ratio of 0.76. That composite included kidney failure, a major loss of kidney function, or death from kidney or cardiovascular causes. Calling this “24% less kidney failure” would misdescribe the endpoint. (Perkovic et al., 2024)

Other trials extend the picture, with different levels of clinical relevance:

Obesity and heart failure with preserved ejection fraction. Tirzepatide reduced worsening heart-failure events and improved health status. The composite benefit was driven by fewer worsening events; cardiovascular mortality benefit was not established. (SUMMIT)

Obesity and moderate-to-severe obstructive sleep apnea. Tirzepatide reduced breathing interruptions and improved sleep-related outcomes over 52 weeks. This does not establish treatment for every cause of poor sleep. (SURMOUNT-OSA)

MASH (metabolic fatty-liver disease with inflammation) and moderate-to-advanced fibrosis. Semaglutide improved liver-biopsy endpoints in the 72-week ESSENCE interim analysis. Histological improvement is not yet proof of fewer liver transplants or liver-related deaths. (ESSENCE)

What the New Tirzepatide Study Adds

The August 2026 BMJ cohort compared 52,971 adults with diabetes and atherosclerotic cardiovascular disease starting tirzepatide or sitagliptin. Over up to one year, tirzepatide was associated with a lower hazard of heart attack, stroke, or death from any cause: 0.68 (95% CI 0.58–0.80). The estimated absolute risk difference was 1.4 percentage points. Sitagliptin served as a cardiovascular-neutral comparison, not an actual placebo. Weighting reduces measured differences, but residual confounding remains; this endpoint also includes noncardiovascular deaths. (Krüger et al., 2026)

The randomized SURPASS-CVOT comparison provides a different answer: against dulaglutide, an established active treatment, tirzepatide met the trial's noninferiority criterion but did not establish superiority for cardiovascular death, heart attack, or stroke. The hazard ratio was 0.92 (95.3% CI 0.83–1.01). A favorable observational comparison against sitagliptin cannot overturn that result; the comparators and questions differ. (Nicholls et al., 2025)

The Infection Signal

There is a plausible reason to investigate infections alongside metabolic disease. A 2026 Lancet analysis of Finnish cohorts and UK Biobank, involving more than 540,000 people, associated obesity with higher rates of severe infections. Its global burden estimates depend on assuming that the associations are causal. It was not a treatment trial: it cannot tell us how much infection risk a GLP-1 medicine would remove, or whether any effect would come from weight loss or another mechanism. (Nyberg et al., 2026)

A systematic review of 136 randomized trials involving 164,322 participants found fewer serious infections with GLP-1-based treatment: relative risk 0.89 (95% CI 0.86–0.93). The reported absolute difference was about three fewer serious infections per 1,000 person-years. Infections were captured as adverse events, and definitions and reporting varied. Several infection categories showed no significant difference. Trial-level links with weight loss and glucose improvement suggest explanations but do not prove mediation in individual patients. (Han et al., 2025)

The JAMA Network Open umbrella review is a useful check on enthusiasm. It examined 60 meta-analyses and found infection and respiratory signals alongside particularly consistent gastrointestinal harms. Many outcomes were incidentally recorded adverse events; studies overlapped across reviews. Its large aggregate participant count therefore should not be read as millions of independent confirmations. The authors called for dedicated confirmation of the proposed benefits. This review also includes Han's infection analysis, so these are partly the same evidence viewed twice. (Yang et al., 2026)

COVID-19: Severity Versus Infection

SELECT's secondary analysis found 43 COVID-19 deaths with semaglutide versus 65 with placebo, with a hazard ratio of 0.66 (95% CI 0.44–0.96). It did not find fewer incident COVID-19 infections. This distinction matters: reduced severity after infection is different from preventing infection. The mortality analysis is informative, but it was not a dedicated trial of treatment started for COVID-19. (Scirica et al., 2024)

The July 2026 Scientific Reports paper adds an observational result. In US insurance data from 2021–2022, semaglutide use was associated with a 49% lower adjusted hazard of COVID-19 death (HR 0.51; 95% CI 0.43–0.61). Age, diabetes, and obesity remained imbalanced after weighting and required further adjustment; dose and average exposure duration were unavailable. Its sensitivity analysis cannot eliminate these limitations. The result supports further study, not prescribing semaglutide to prevent or treat COVID-19. (Tang et al., 2026)

Where the Evidence Is Mixed

Alcohol use disorder is a promising research direction. A 2026 randomized trial enrolled 108 people with alcohol use disorder and obesity; both groups received cognitive behavioral therapy. Over 26 weeks, semaglutide produced a larger reduction in heavy-drinking days than placebo. This is stronger evidence than prescription-database associations, but a small, single-center trial does not establish long-term effectiveness or applicability to people without obesity. (Klausen et al., 2026)

Alzheimer's disease illustrates why plausible mechanisms need clinical testing. The evoke and evoke+ trials randomized 3,808 people with early symptomatic Alzheimer's disease. Oral semaglutide did not slow clinical progression at 104 weeks. These negative treatment results do not settle whether metabolic treatment could influence future dementia risk, but they directly challenge claims that semaglutide is an established Alzheimer's therapy. (Cummings et al., 2026)

Risks That Matter

Nausea, vomiting, diarrhea, and constipation are common, and tolerability can determine whether treatment continues. Severe gastrointestinal symptoms can cause dehydration and acute kidney injury. Wegovy is not recommended in severe gastroparesis. Its label calls for monitoring in diabetic retinopathy and warns about hypoglycemia with insulin or insulin-releasing medicines. For weight management, it specifies stopping when pregnancy is recognized and at least two months before a planned pregnancy. (Wegovy prescribing information, 2026)

Gallbladder and biliary disease are a recognized concern in randomized-trial evidence. Pancreatitis requires a separate distinction: uncertainty about the size of an excess risk does not justify ignoring a potentially serious event. In January 2026, the UK MHRA strengthened warnings about acute pancreatitis, including rare severe and fatal cases. Persistent severe abdominal pain, sometimes spreading to the back, warrants urgent medical assessment. (He et al., 2022; MHRA pancreatitis update, 2026)

The EMA classifies NAION, an optic-nerve condition that can cause vision loss, as a very rare semaglutide side effect. Epidemiological evidence suggested roughly one additional case per 10,000 person-years of treatment. Sudden loss of vision or rapidly worsening eyesight needs prompt medical attention. This is a semaglutide-specific regulatory conclusion, not proof of the same risk for every medicine in the group. (EMA NAION review, 2025)

The US Wegovy and Zepbound labels contraindicate use with a personal or family history of medullary thyroid carcinoma or MEN2; the tumor warnings originate from animal findings whose human relevance is unknown. Tirzepatide can reduce oral contraceptive effectiveness: its label specifies non-oral or additional barrier contraception for four weeks after starting and after each dose increase. Delayed stomach emptying also matters before anesthesia or deep sedation; the procedural team needs to know about treatment. (Wegovy prescribing information, 2026; Zepbound prescribing information, 2026)

Safety assessment also means retiring unsupported concerns. In January 2026, the FDA requested removal of suicidal-behavior and ideation warnings from affected GLP-1 medicine labels after a comprehensive review found no increased risk. That is reassuring evidence about a specific proposed harm, not a guarantee against every psychiatric symptom. (FDA safety communication, 2026)

Treatment Beyond the First Year

Weight reduction includes some lean tissue. In the SURMOUNT-1 body-composition substudy, tirzepatide reduced both fat and lean mass. Lean mass is not synonymous with skeletal muscle or strength, but the finding makes nutritional adequacy and physical function relevant, especially for vulnerable older adults. A joint professional advisory emphasizes adequate nutrition and resistance training during treatment; the scale alone is an incomplete measure of success. (Look et al., 2025; Joint nutrition advisory, 2025)

Stopping treatment is another practical issue. In the STEP 1 extension, participants regained about two-thirds of their previous weight loss within a year after semaglutide and the study's lifestyle intervention ended. Many cardiometabolic measures moved back toward baseline. This does not mean everyone regains the same amount, but it argues against treating a short course as a permanent metabolic reset. (Wilding et al., 2022)

How Strong Is the Evidence?

These are editorial judgments about the evidence discussed above, rather than a single rating for every GLP-1 medicine or use. Randomized clinical outcomes, exploratory infection analyses, and safety surveillance answer different questions. (FLOW; Yang et al.; EMA)

Dimension	Assessment
Clinical relevance	High for cardiovascular and kidney events in the studied populations; Moderate for newer uses that rely on surrogate or secondary outcomes.
Study quality	High for the major randomized outcome trials; Moderate for infection prevention, where overlapping reviews and adverse-event reporting limit interpretation.
Safety data	Moderate overall: common harms are well characterized, while rare events and prolonged use in less-studied populations require continued surveillance.

The strongest case is disease-specific: meaningful benefits in selected people, with meaningful tradeoffs. Fewer deaths in a high-risk clinical trial are not proof that a drug slows biological aging or extends life in healthy people. The research supports taking these medicines seriously without turning emerging associations into a general longevity prescription. (SELECT mortality analysis; SURPASS-CVOT; MHRA patient guidance)

Header and thumbnail: AI-generated editorial illustrations of an unbranded injection pen, not photographs of a specific medicine.