What Do GLP-1 Drugs Do Beyond Weight? — Signal, Association, and Confirmation

Conflicting News About the Same Drug

Records from a university hospital neuro-ophthalmology clinic analyzed that optic neuropathy occurred much more frequently in people prescribed semaglutide. The risk was 4.28 times higher in patients with type 2 diabetes and 7.64 times higher in people who were overweight or obese. Because it was an observational study with a small number of events, the authors wrote that more research is needed to determine causality (PMID 38958939). Later, an analysis pooling the records of 37.1 million diabetes patients yielded a much smaller increase in risk for the same question (PMID 39976940). Given the same drug and the same condition, the numbers diverge to this extent.

GLP-1 receptor agonists are drugs made to bind to GLP-1 receptors and send signals just like GLP-1. Semaglutide belongs to this category, and tirzepatide is a drug made to bind simultaneously to the receptors of two hormones released from the gut, GIP and GLP-1. This article refers to drugs in this class as "GLP-1 drugs" and mainly covers semaglutide and tirzepatide.

How these drugs reduce appetite and lower weight, common gastrointestinal symptoms and gallstones, concerns about muscle loss, and the return of weight after stopping are covered in "What Are Obesity and GLP-1 Drugs? — A Letter That Lingers, a Weight That Wants to Return". A trial showing reduced cardiovascular events in adults with obesity or overweight who already had cardiovascular disease and did not have diabetes is also in that article (PMID 37952131). This article follows the next question. What are these drugs doing in the body besides weight loss, and how much of that has been confirmed?

Stories circulating about these drugs fall into three branches. There are stories of weight loss injections, expectations that they are cure-alls that even treat dementia and addiction, and warnings that they harm the eyes and mind and cause cancer. Each of the three stories cites papers as evidence. A significant portion of the contradictions is resolved by looking at which tier and from whom that evidence came.

Three Layers of Evidence

If a person taking a medication experiences sudden blurred vision, either the doctor, the patient, or the pharmaceutical company can report the event to the authorities. The first layer is this report. Pharmacovigilance is the work of gathering adverse event reports after a marketed drug is used to find rare or unexpected harms. The FDA's Adverse Event Reporting System is a representative reporting channel.

There is a way to look for alarms within a pile of reports. Disproportionality analysis is a method of measuring whether a specific drug and a specific adverse event have been reported together unusually more often than other drugs within reporting data. Reports do not contain the total number of people who took that drug. What disproportionality analysis gives is a signal, and it cannot speak to causality (PMID 40133108). To measure what the actual risk is, a prospective study that verifies diagnoses is needed separately (PMID 42259643).

The second layer is observational studies. Observational studies are studies that do not interfere with people's lives, leaving them as they originally were and watching what happens afterwards. They compare people who used this drug and people who used another drug in medical records and insurance claims data. The weakness is that the two groups may differ from the beginning. A confounding factor is a third factor that makes people who used the drug and those who did not different from the start, creating instead a difference that looks like the drug's effect. A design that sets the comparison target as another drug used for the same disease is a method to reduce this problem (PMID 40133108, PMID 39226030).

The third layer is the layer of confirmation. A randomized controlled trial is a trial that randomly divides participants like a drawing of lots, gives the drug to only one side, and compares the results with the other side. The other side is commonly given a placebo. A placebo is a fake medicine that looks and is used the same as the drug, but has no medicinal ingredients. Entrusting the division to a drawing of lots reduces the room for the two groups to differ from the start.

A review paper that organizes these layers regarding thyroid cancer distinguishes them like this. Randomized trials provide the strongest evidence for causality, observational studies look at real-world results in a larger population but cannot account for all other factors, and pharmacovigilance studies signal alerts but do not address causality (PMID 38343381). Randomized trials also have blind spots. Trials are too small and short to catch rare diseases. An analysis combining 48 trials looking at cancer noted the limitation that those trials were not originally designed to look at cancer and had short follow-ups (PMID 41359966).

Reporting data gives signals, observational studies give associations, and randomized trials give confirmation. When news about the same drug is mixed, the first question to ask is which layer that news came from.

What the Trials Confirmed

3,533 people with type 2 diabetes and pre-existing kidney damage causing protein to leak into their urine received semaglutide or a placebo once a week. An interim analysis recommended stopping the trial early, and at about 3 years of follow-up, the proportion of people who lost kidney function (such as through dialysis or transplantation), whose kidney function measure dropped by half or more from baseline, or who died from kidney or cardiovascular causes was 19% in the semaglutide group and 23% in the placebo group (PMID 38785209, PMID 41706532). Death from any cause was also 13% and 16% (PMID 41706532). An analysis combining 11 large-scale trials also showed that kidney events in type 2 diabetes patients were reduced by 18% with this class of drugs, and discontinuing the drug due to side effects was more frequent in the drug group (PMID 39608381).

On the cardiovascular side, one type of heart failure was put to the test. Heart failure is a condition where the heart either fails to pump as much blood as the body needs, or can only do so by raising the filling pressure. The ejection fraction is the percentage of blood in the ventricle that is pumped out during a single contraction. Heart failure with preserved ejection fraction is heart failure that occurs even though the ejection fraction is maintained at 50% or higher. Around the time the first trial paper below was published, no treatments had been approved targeting this heart failure that accompanies obesity (PMID 37622681).

In a 52-week trial of 529 patients with obesity and heart failure with preserved ejection fraction, the symptom and physical limitation score (0 to 100 points, higher is better) increased by 16.6 points in the semaglutide group and 8.7 points in the placebo group, and the distance walked in 6 minutes increased by 21.5m and 1.2m (PMID 37622681). In a trial where 731 patients under the same conditions received tirzepatide or a placebo and were followed for around 2 years, cardiovascular death or worsening heart failure occurred in 9.9% and 15.3% (PMID 39555826). The reduced portion came from worsening heart failure events (8.0% vs. 14.2%). Looking at cardiovascular death alone, the events were few at 8 patients versus 5, making it impossible to determine a difference. Symptom and physical limitation scores also improved more than with placebo, and people who stopped the drug due to side effects (mainly gastrointestinal symptoms) were 6.3% and 1.4% (PMID 39555826). In a pooled post-hoc analysis of 3,743 patients with heart failure with preserved or mildly reduced ejection fraction across four semaglutide trials, cardiovascular death or worsening heart failure was also reduced to 5.4% and 7.5%, but there was no meaningful difference for cardiovascular death alone (PMID 39222642).

On the respiratory side, there is sleep apnea. Obstructive sleep apnea is a condition in which the airway at the back of the throat repeatedly narrows or closes during sleep, causing breathing to be greatly reduced or to stop. The apnea-hypopnea index is the number of times breathing stops or is significantly reduced per hour of sleep. There are two phase 3 trials in adults with moderate-to-severe sleep apnea and obesity. One recruited people who do not use a continuous positive airway pressure machine during sleep, and the other recruited those who do. This index was around 50 times per hour at baseline, and over 52 weeks, it decreased by 25.3 and 29.3 times in the tirzepatide groups, and by 5.3 and 5.5 times in the placebo groups (PMID 38912654).

There are also liver trials. Metabolic dysfunction-associated steatohepatitis is a condition where, in addition to fat accumulation in the liver, inflammation and liver cell damage occur, which can progress to fibrosis, where the liver hardens like a scar. In a phase 3 trial of 1,197 people with biopsy-confirmed steatohepatitis and moderate-to-advanced fibrosis, an interim analysis of 800 patients who completed 72 weeks first showed that steatohepatitis resolved without worsening fibrosis in 62.9% of the semaglutide group and 34.3% of the placebo group (PMID 40305708). Fibrosis improved by one stage or more without worsening steatohepatitis in 36.8% and 22.4% (PMID 40305708). Based on these interim results, accelerated approval was granted in the United States, but formal approval awaits long-term results, and cirrhosis is outside the scope of approval (PMID 41201884). In a trial evaluating patients with cirrhosis, neither resolution of steatohepatitis nor improvement in fibrosis differed from placebo (PMID 40736113). An analysis combining the trials noted that while this class of drugs is a promising treatment option, the long-term effects on clinical events originating in the liver require further study (PMID 40736113).

There are also the knees and legs. In a 68-week trial of 407 people with obesity, moderate knee osteoarthritis, and moderate-to-severe pain, the knee pain score (0 to 100 points, higher is worse) decreased by 41.7 points in the semaglutide group and 27.5 points in the placebo group, and body weight decreased by 13.7% and 3.2%. Both groups received diet and exercise counseling together (PMID 39476339). In a 52-week trial of 792 people with type 2 diabetes and peripheral artery disease who experience leg pain when walking, the maximum walking distance became 1.21 times the baseline in the semaglutide group and 1.08 times in the placebo group (PMID 40169145). The authors noted that the effect in patients with peripheral artery disease without diabetes should be tested separately (PMID 40169145).

All trials in this section were funded by the companies that manufactured the drugs (PMID 38785209, PMID 37622681, PMID 39555826, PMID 38912654, PMID 40305708, PMID 39476339, PMID 40169145). The confirmed benefits all belong to specific populations: people with both diabetes and kidney disease; patients with obesity and heart failure with preserved ejection fraction, sleep apnea, or knee osteoarthritis; patients with biopsy-confirmed steatohepatitis and moderate-to-advanced fibrosis (interim analysis); and patients with diabetes and peripheral artery disease.

Is It Because of Weight Loss?

In the heart failure trial of tirzepatide, over 52 weeks, body weight decreased by 10.4% in people with type 2 diabetes and by 12.9% in those without. However, the extent to which worsening heart failure events decreased and symptoms, walking distance, and quality of life improved were similar in both groups, as was the extent to which the fat around the heart and the weight of the heart muscle decreased (PMID 40903131). The authors concluded that patients with diabetes showed a good response even though they lost less weight, and noted that this result raises the possibility that the heart failure benefits of this drug class cannot be properly gauged by the magnitude of weight change (PMID 40903131).

In another analysis of the same trial, compared to placebo over 52 weeks, estimated blood volume decreased by 0.58 liters, systolic blood pressure dropped by 5 mmHg, and inflammatory markers were lowered by 37% (PMID 39551891). A relationship was seen where symptoms and walking distance improved as much as blood volume decreased, but this correlation was a post-hoc exploratory analysis (PMID 39551891). There are observations pointing the other way as well. In a cardiac imaging substudy, the weight of the heart muscle decreased by 11g compared to placebo, and that change went hand in hand with weight loss (PMID 39566869).

One of the mechanisms discussed beyond weight loss is the action of suppressing inflammation. This action has been reported across various tissues in test-tube and animal experiments, and is used as a basis for finding new indications (PMID 35738455). Because this is evidence from the test-tube and animal stage, it must be tested separately to see what benefits it translates to in humans. Even in people who lost less weight, the heart failure benefits were similar, and the authors raised the possibility that these benefits cannot be gauged solely by the magnitude of weight change. What plays a role other than weight remains a hypothesis. The hypothesis that the brain can be protected through various pathways, including the anti-inflammatory action, was the starting point for the trials in the next section (PMID 39780249).

What Observation Expected and Trials Did Not Confirm

In records of people with diabetes or obesity, those who used this class of drugs showed a lower risk of dementia and Alzheimer's disease, and animal experiments pointed the same way (PMID 41865758). However, some point out that recent animal research and the fact that this class of drugs does not enter the brain well run counter to the interpretation that it directly protects the brain (PMID 42580680). There were also observational data showing less dementia compared to people who used other diabetes drugs (PMID 42580680). This expectation born at the level of association was brought to the level of confirmation by two Phase 3 trials named evoke and evoke+.

The participants were 55 to 85 years old, had amyloid—a protein that accumulates in the brain in Alzheimer's disease—confirmed by testing, and had mild cognitive impairment or mild dementia as a result. Across multiple countries, 3,808 people received once-daily oral semaglutide or a placebo (PMID 41865758). A primary endpoint is a measurement set in advance before starting a trial to decide success or failure. When there are multiple scores and time points, a difference can happen by chance in just one of them, so the trial's success or failure is decided by this value. Both trials set as their primary endpoint how much a score summing dementia symptoms across several domains such as memory, judgment, and daily living (higher means worse) increased over 104 weeks, and an extension period of 52 weeks was planned in advance after that (PMID 39780249).

Over 104 weeks, this score increased by 2.3 points in the semaglutide group and 2.3 points in the placebo group in evoke, and by 2.2 and 2.1 points in evoke+ (PMID 41865758). Both trials showed no difference and were stopped for negative results. The authors concluded that oral semaglutide did not slow the clinical progression of early Alzheimer's disease, and safety was consistent with what has been seen in other conditions (PMID 41865758). The trials were funded by the company that makes the drug (PMID 41865758).

After the results came out, a debate ensued. In a separately published post-hoc reanalysis, one researcher recalculated the 130-week and 156-week time points using the published means and errors, and argued that evoke's 130-week daily living score showed a meaningful difference and other cognitive scores also showed a trend favoring the drug. He traced the cause of the failure to semaglutide not easily crossing from the blood into the brain (PMID 42261705). Authors of a rebuttal countered that both trials missed both the primary endpoint and secondary endpoints regarding cognition and function, and that participants dropped out during the extension period so that the weight of inference from the remaining people alone is small (PMID 42720565). Interpretations also diverge regarding cerebrospinal fluid biomarkers. The researcher who published the reanalysis saw meaningful differences in some markers (PMID 42261705), and one review summarized that there were favorable changes in biomarkers (PMID 42580680), but the authors of the rebuttal wrote that reading the sub-studies in full does not neatly establish the interpretation that the disease's progression was altered (PMID 42720565).

The scope of what these results say is narrow. The participants were patients with early symptomatic Alzheimer's disease whose symptoms had already appeared, and the drug was oral semaglutide. The authors of a review wrote that while it showed no effect in symptomatic Alzheimer's disease, its effects on metabolism could be beneficial at an earlier time like the pre-symptomatic stage, and future research must figure out whether starting early in a group with high metabolic risk changes the course (PMID 42580680). The authors of the rebuttal also wrote that evoke did not refute the cardio-metabolic hypothesis, and that the question has shifted to which pathways, which patients, and which disease stages will benefit (PMID 42720565).

The lower dementia risk shown by observational studies did not translate into an effect of slowing progression in two Phase 3 trials involving patients with early symptomatic Alzheimer's disease with confirmed amyloid.

What Is Still Being Tested

The hope for alcohol comes from animal experiments and the observation that people taking these drugs want less alcohol (PMID 38806481, PMID 42070571). In medical records of 83,825 people with obesity, those prescribed semaglutide had a 50 to 56% lower risk of developing or relapsing into alcohol use disorder over one year than those who used other obesity drugs (PMID 38806481). In records of about 520,000 U.S. veterans with type 2 diabetes, those who started other classes of diabetes drugs also showed a lower risk of developing use disorders for alcohol, cannabis, cocaine, nicotine, and opioids compared to those who started other drugs (PMID 41781010). The authors wrote that these observational data are worth evaluating further (PMID 41781010).

Randomized trials are still small, and results vary by trial and measure. In a 9-week trial of 48 adults with alcohol use disorder who did not seek treatment, semaglutide reduced the amount of alcohol consumed in a laboratory setting, the amount drunk on drinking days, and craving, but did not change the average daily alcohol consumption or the number of drinking days (PMID 39937469). In a single-center trial of 108 treatment-seeking individuals with both obesity and moderate or severe alcohol use disorder who received injections for 26 weeks alongside cognitive behavioral therapy, the primary endpoint of the percentage of heavy drinking days decreased by 41.1 percentage points in the drug group and 26.4 percentage points in the placebo group, showing a meaningful difference. The authors concluded that it showed a clear therapeutic effect (PMID 42070571). In a trial giving oral semaglutide for 8 weeks to 50 treatment-seeking individuals, the primary endpoint of laboratory craving did not decrease and daily alcohol intake remained unchanged, but heavy drinking days, alcohol intake on drinking days, and everyday cravings decreased. The authors reported confirming the results of earlier trials and wrote that it is worth continuing development (PMID 42522065).

Depending on how these data are grouped, the conclusions also change. An analysis pooling only three randomized trials (430 people) found that the reductions in alcohol consumption and craving were not statistically meaningful, whereas observational studies showed a reduction in alcohol-related events (PMID 41350683). Another analysis combining observational studies and trials concluded that drinking decreased, but there was high variability between studies (PMID 41324012). The authors of both analyses wrote that larger randomized trials are needed (PMID 41350683, PMID 41324012).

In Parkinson's disease, there is a phase 2 trial of lixisenatide, another drug in the same class. In 156 early-stage patients within 3 years of diagnosis, after 12 months, the motor disability score worsened by 3.04 points in the placebo group and remained almost unchanged in the drug group (PMID 38598572). 46% of the drug group experienced nausea, and the authors concluded that longer and larger trials are needed (PMID 38598572). An analysis pooling 7 trials (973 people) of diabetes drugs including GLP-1 drugs for Parkinson's disease showed no meaningful difference from placebo in motor symptom scores. On cognitive and depression scales, the GLP-1 drug group showed signs of being better, while gastrointestinal side effects increased (PMID 40762930).

For alcohol, observational data and small trials support expectations, but results vary by trial and measure; for Parkinson's disease, one small trial of lixisenatide showed a favorable direction, while an analysis pooling trials showed no difference in motor symptoms. Both fields are awaiting answers from larger trials.

Two Layers Diverging in the Eye

The 4.28-fold and 7.64-fold figures at the beginning are numbers that came out regarding a disease of the optic nerve. NAION (non-arteritic anterior ischemic optic neuropathy) is a disease where blood does not circulate well to the front end of the optic nerve that sends visual signals from the eye to the brain, usually causing the vision in one eye to suddenly drop. If you follow the path that this signal has walked, three layers appear in sequence.

The starting point was the records of a university hospital's neuro-ophthalmology clinic. A review revisiting this analysis pointed out that the incidence of NAION even in people not taking semaglutide was orders of magnitude higher than in the general population. The interpretation is that people at high risk flocked in due to the pathway through which patients were referred (PMID 40055951). In U.S. reporting data up to the end of 2024, a disproportionality analysis caught a signal only for semaglutide among GLP-1 drugs, and that signal was maintained even when changing the comparison drug. Reports surged to 53 cases over the recent three months, with 18 of them coming from Denmark (PMID 40133108).

At the layer of association, the picture became smaller and diverged. An analysis gathering records from medical institutions in multiple countries showed no association between semaglutide and NAION (PMID 39491755). In another analysis combining multiple databases, the incidence of NAION in type 2 diabetes patients using semaglutide was 14.5 cases per 100,000 person-years, and while there was an increased risk, it was smaller than the initial report (PMID 39976940). Among four large observational studies summarized by a review, two reported a 2- to 3-fold higher incidence, one reported a trend only in diabetes patients, and one reported no meaningful difference, while an analysis pooling the trials showed no difference (PMID 40055951).

The European Medicines Agency classified NAION as a very rare side effect of semaglutide (PMID 41857480). In an analysis up to the third quarter of 2025, with more reports accumulated, signals were caught for semaglutide, tirzepatide, and liraglutide, but not for dulaglutide, exenatide, or lixisenatide. The authors noted that this difference argues against the interpretation that the effect is shared across the entire class, and that prospective studies with a neuro-ophthalmologist confirming the diagnosis are needed to establish causality and absolute risk (PMID 42259643). There is a hypothesis that blood flow in the optic nerve head changes as blood sugar drops rapidly, and a hypothesis that the risk is greater in eyes where the optic nerve head is small and crowded. The author of the review concluded that because the data are observational, causality cannot be inferred (PMID 40055951).

In the same eye, there is also a risk that increased in randomized trials. Diabetic retinopathy is a disease where the small blood vessels of the retina—the membrane that receives light inside the back of the eye—are damaged. In a 2-year trial of 3,297 type 2 diabetes patients at high cardiovascular risk, retinopathy complications requiring vitreous hemorrhage, blindness, or injections or laser treatment were meaningfully higher in the semaglutide group at 1.76 times that of the placebo group (PMID 27633186). One review summarized that this worsening was concentrated in patients who already had retinopathy and whose blood sugar dropped rapidly, while two large trials of other GLP-1 drugs showed no meaningful increase, suggesting it may differ by drug (PMID 41857480). One review cited a strong blood sugar-lowering action as the background, and wrote that patients at risk of worsening existing retinopathy, especially those using insulin together, should be watched carefully (PMID 34305810).

The optic nerve signal has become smaller than the initial report but has not disappeared, and causality and absolute risk have not yet been established. Retinopathy complications increased in one semaglutide trial in type 2 diabetes patients with high cardiovascular risk, and did not increase in two trials of other GLP-1 drugs.

Mind, Cancer, Pancreas, and Anesthesia

Reports of suicidal thoughts continued to pour in, prompting European regulators to launch an investigation (PMID 38182782). In an analysis pooling 27 randomized trials (about 32,354 participants), suicide- and self-harm-related events were rare and showed no difference, at roughly 0.047 cases per 100 person-years in the drug group and 0.042 cases in the placebo group (PMID 40105856). A post-hoc pooled analysis of semaglutide trials in adults with obesity or overweight without a known major psychiatric disorder similarly showed no increase in depressive symptoms or suicidal thoughts (PMID 39226070).

At the observational level, results diverge. An analysis comparing people who started a different class of diabetes drug in Sweden and Denmark (mostly patients with type 2 diabetes) found no increase in suicide death, self-harm, depression, and anxiety, estimating the upper limit for a potential increase in suicide death at 0.16 cases per 1,000 person-years (PMID 39226030). An analysis from a medical records network comparing semaglutide with other obesity and diabetes drugs actually showed a lower risk of suicidal thoughts over 6 months, and the authors noted that their findings do not support the interpretation that the risk is higher (PMID 38182782). Conversely, a large cohort matched with non-users with obesity reported a 195% higher risk of major depression and a 106% higher risk of suicidal behavior, prompting the authors to write that clinicians should carefully review patients' medical history before prescribing and that prospective trials are needed (PMID 39424950). An analysis combining trial, observational, and reporting data found no meaningful increase, yet noted that this result should be read as failing to detect an increased risk rather than serving as definitive proof that no risk exists (PMID 42584177). Another analysis pooling observational studies found no significant association with suicidal thoughts or behaviors, but reported that high heterogeneity between studies calls for caution and that clinicians should monitor patients, particularly those with psychiatric conditions (PMID 39945396).

It is not just the active ingredient that amplifies signals in reporting data. Compounded GLP-1 products—prepared separately instead of approved finished products—do not undergo the same testing as approved drugs. In US reporting data, these compounded products showed an unusually high concentration of suicide-related and cholecystitis reports compared to approved products (disproportionality scores of 6.34 and 3.39), alongside many reports of compounding errors and contamination (PMID 40285721). The authors noted that this analysis was not designed to show causality (PMID 40285721).

The thyroid cancer warning originated from rodent experiments. The warning label was attached to this class of drugs because rodents developed medullary thyroid cancer, a rare form of the disease, but the relationship with other thyroid cancers in humans remains debated. Synthesizing these findings, this review concluded that there is no definitive evidence of increased risk, while also noting that unfounded worries can lead to unnecessary thyroid screenings and overdiagnosis (PMID 38343381). Analyses pooling trials are mixed. One analysis yielded a 1.52-fold thyroid cancer risk, but the statistical significance would vanish if just a single event changed, and the magnitude meant that 1,349 people would need to use the drug for 5 years for one extra case to occur (PMID 38018310). A more recent analysis pooling 48 trials (94,245 participants) found little to no effect on thyroid cancer (PMID 41359966). Population records from six countries showed no increased risk in patients with type 2 diabetes, though the follow-up of 2 to 3 years precluded ruling out long-term risks (PMID 39772758), and records of patients with type 2 diabetes using the drug for over 1 year and followed for a median of 4.5 years similarly showed no increase (PMID 41182904).

For cancer, signals pointing in the opposite direction also exist. Records of 1.1 million individuals with obesity showed that those using the drug had lower risks of gastrointestinal and breast cancers, though liraglutide within the same data showed a higher risk of thyroid cancer (PMID 39796706). Another analysis pooling 50 trials found no difference in overall cancer and most specific cancers. Within those, uterine cancer decreased in trials targeting obesity, while the thyroid cancer risk signal was more prominent in longer trials. Colorectal cancer increased slightly in short trials, which the authors attributed to patients undergoing more diagnostic workups triggered by gastrointestinal symptoms (PMID 40437949).

On the pancreatic side, reporting data and trials tell different stories. The association with pancreatitis has been raised primarily through pharmacovigilance and case reports (PMID 41351656), and post-marketing reporting analyses of semaglutide have also captured unexpected signals such as pancreatic cancer, intestinal obstruction, and cholecystitis (PMID 38943656). However, in an analysis pooling 21 trials, the incidence of pancreatitis and tumors overall did not differ from the control group (pancreatic cancer was not analyzed separately), whereas gastrointestinal disorders and gallbladder disease increased (PMID 40892610).

Anesthesia-related concerns stem from the stomach. Aspiration refers to food, water, or saliva entering the airway below the vocal cords. There is concern that these drugs slow down the rate at which the stomach empties, potentially allowing residual contents in the stomach to cross into the airway during anesthesia (PMID 39418085). An analysis pooling observational studies found that people using these drugs were more likely to have residual gastric contents despite adhering to fasting instructions, but no association with aspiration emerged. The certainty of evidence for both conclusions is low (PMID 40230298). It is concluded that evidence is insufficient to establish recommendations on how to manage these drugs before surgery or endoscopy (PMID 39418085, PMID 39096914).

Safety data are particularly sparse for certain groups. Safety data are limited for pregnant and breastfeeding women, children, and individuals with severely impaired kidney or liver function (PMID 41351656). While there are cases of unintended pregnancies while using the drug, evidence to predict the impact of exposure around the time of conception is lacking. Summarizing human data to date, this review noted that no consistent pattern of congenital malformations has emerged, yet it still took the stance of recommending contraception while taking the drug due to this gap in evidence (PMID 39181497).

Most concerns regarding the mind, cancer, pancreas, and anesthesia originate at the level of signals and associations, pooled trial analyses showed no difference for suicide and self-harm, and some findings point in dual directions. "There is no evidence of an increase" and "proof that no risk exists" are two different things.

When Reading a Single Piece of News

The benefits for specific groups, such as the kidneys, heart failure, and sleep apnea, have been confirmed in trials. There are cases where expectations raised by observational studies were not confirmed in large-scale trials, such as in early Alzheimer's disease. There are also cases like the optic nerve where signals remained diminished after passing through massive records. Because the evidence surrounding a single drug stands on these three diverging paths at the same time, articles pointing in opposite directions can come out in the same week.

When reading a single piece of news, there are five things you can ask. They are whether that number came from the layer of reporting, observation, or trial; who the subjects were (whether they had diabetes, which stage the disease was at); what drug and what product it was; what it was compared against (a placebo, another drug, or people not taking the drug); and what the absolute number was behind the phrase "so many folds." In the optic nerve story from earlier, 4.28 times is the relative risk comparing patients at a single hospital, and 14.5 cases per 100,000 person-years is the incidence rate from among tens of millions of records for patients with type 2 diabetes who newly took semaglutide.

In the populations confirmed by trials, this drug improved outcomes for the kidneys, heart, breathing, liver, joints, and legs in addition to weight. Whether that goes separately from weight loss is still at the possibility stage, and many other expectations and worries are still waiting for the answers of trials. The principles of reducing weight and what happens after stopping are covered in "What Are Obesity and GLP-1 Drugs? — A Letter That Lingers, a Weight That Wants to Return".

Source Papers

• "Risk of Nonarteritic Anterior Ischemic Optic Neuropathy in Patients Prescribed Semaglutide." (2024) — PMID 38958939 — Retrospective matched cohort from an academic neuro-ophthalmology registry, NAION risk in semaglutide-prescribed group type 2 diabetes HR 4.28 and overweight/obesity HR 7.64, observational study so causality requires further research.
• "Semaglutide and Nonarteritic Anterior Ischemic Optic Neuropathy." (2025) — PMID 39976940 — OHDSI 14 databases and 37.1 million type 2 diabetes patients, 14.5 NAION cases per 100,000 person-years in semaglutide users, a modest risk increase smaller than initial reports.
• "Semaglutide and Cardiovascular Outcomes in Obesity without Diabetes." (2023) — PMID 37952131 — SELECT, major adverse cardiovascular events 6.5% vs 8.0% in overweight/obesity adults with cardiovascular disease without diabetes.
• "Disproportionality analysis on semaglutide and nonarteritic anterior ischemic optic neuropathy in the FDA adverse event reporting system: An emerging pharmacovigilance signal?" (2025) — PMID 40133108 — FAERS disproportionality analysis, signal only in semaglutide and persists in active comparator group, 53 cases in the last 3 months (18 in Denmark), causal inference not possible.
• "Optic Ischaemic Neuropathy in Incretin-Based Therapy: A Comparative Analysis of Real-World Safety Data." (2026) — PMID 42259643 — FAERS, semaglutide, tirzepatide, and liraglutide signals, no signal for dulaglutide, exenatide, and lixisenatide, contradicts class-wide effect, prospective study with confirmed diagnosis needed.
• "GLP-1 Receptor Agonist Use and Risk of Suicide Death." (2024) — PMID 39226030 — Swedish and Danish active-comparator new-user cohort (mostly type 2 diabetes), no increase in suicide death, self-harm, or depression/anxiety, absolute increase upper bound 0.16 cases per 1,000 person-years.
• "Glucagon-Like Peptide-1 Receptor Agonists and Thyroid Cancer: A Narrative Review." (2024) — PMID 38343381 — Characterization of trial, observational, and pharmacovigilance evidence, rodent medullary carcinoma leading to warnings, human non-medullary relationship debated, no definitive evidence.
• "Risk for Cancer With Glucagon-Like Peptide-1 Receptor Agonists and Dual Agonists : A Systematic Review and Meta-analysis." (2026) — PMID 41359966 — 48 RCTs and 94,245 individuals, little to no effect on obesity-related cancers including thyroid cancer (moderate certainty), trials not designed to look at cancer with short follow-up.
• "Effects of Semaglutide on Chronic Kidney Disease in Patients with Type 2 Diabetes." (2024) — PMID 38785209 — FLOW, 3,533 type 2 diabetes and chronic kidney disease patients, 24% reduction in major kidney events, early discontinuation recommended at interim analysis, funded by Novo Nordisk.
• "Kidney and Survival Outcomes with Semaglutide by CKD Severity in the FLOW Trial." (2026) — PMID 41706532 — FLOW primary outcome 19% vs 23%, death 13% vs 16%, consistent across kidney disease severity.
• "Effects of GLP-1 receptor agonists on kidney and cardiovascular disease outcomes: a meta-analysis of randomised controlled trials." (2025) — PMID 39608381 — 11 type 2 diabetes RCTs, 18% reduction in composite kidney events, discontinuations due to adverse events more frequent on the drug side.
• "Semaglutide in Patients with Heart Failure with Preserved Ejection Fraction and Obesity." (2023) — PMID 37622681 — STEP-HFpEF, 529 patients for 52 weeks, symptom score 16.6 vs 8.7, 6-minute walk 21.5m vs 1.2m, no approved treatments for obesity-related HFpEF existed.
• "Tirzepatide for Heart Failure with Preserved Ejection Fraction and Obesity." (2025) — PMID 39555826 — SUMMIT 731 patients, cardiovascular death or worsening heart failure 9.9% vs 15.3%, worsening heart failure 8.0% vs 14.2%, cardiovascular death 8 patients vs 5 patients, funded by Eli Lilly.
• "Semaglutide versus placebo in patients with heart failure and mildly reduced or preserved ejection fraction: a pooled analysis of the SELECT, FLOW, STEP-HFpEF, and STEP-HFpEF DM randomised trials." (2024) — PMID 39222642 — Post hoc pooled analysis of 3,743 patients, composite 5.4% vs 7.5%, cardiovascular death alone not significant.
• "Tirzepatide for the Treatment of Obstructive Sleep Apnea and Obesity." (2024) — PMID 38912654 — 2 SURMOUNT-OSA phase 3 trials, baseline AHI around 50, −25.3/−29.3 vs −5.3/−5.5 at 52 weeks, PAP non-users and users separated.
• "Phase 3 Trial of Semaglutide in Metabolic Dysfunction-Associated Steatohepatitis." (2025) — PMID 40305708 — Interim analysis of 800 out of 1,197 patients at 72 weeks in ESSENCE, MASH resolution 62.9% vs 34.3%, fibrosis improvement 36.8% vs 22.4%.
• "Semaglutide therapy for metabolic dysfunction-associated steatohepatitis: November 2025 updates to AASLD Practice Guidance." (2026) — PMID 41201884 — FDA accelerated approval based on intermediate results evidence, formal approval awaits long-term outcomes, cirrhosis outside the approved scope.
• "Glucagon-Like Peptide-1 Receptor Agonists Improve MASH and Liver Fibrosis: A Meta-Analysis of Randomised Controlled Trials." (2025) — PMID 40736113 — 13 RCTs, resolution and fibrosis improvement in F2-F3 MASH, ineffective in 1 compensated cirrhosis trial, long-term effects on liver-related clinical events require further study.
• "Once-Weekly Semaglutide in Persons with Obesity and Knee Osteoarthritis." (2024) — PMID 39476339 — STEP 9, 407 patients for 68 weeks, WOMAC pain −41.7 vs −27.5, body weight −13.7% vs −3.2%.
• "Semaglutide and walking capacity in people with symptomatic peripheral artery disease and type 2 diabetes (STRIDE): a phase 3b, double-blind, randomised, placebo-controlled trial." (2025) — PMID 40169145 — 792 patients for 52 weeks, maximum walking distance ratio 1.21 vs 1.08, studies in patients without diabetes needed.
• "Influence of Type 2 Diabetes on the Effects of Tirzepatide in Patients With Heart Failure and a Preserved Ejection Fraction With Obesity: A Prespecified Stratification-Based Analysis." (2025) — PMID 40903131 — Smaller weight loss with diabetes at 10.4% vs 12.9% but similar heart failure benefits, possibility that weight change cannot gauge benefit.
• "Effects of tirzepatide on circulatory overload and end-organ damage in heart failure with preserved ejection fraction and obesity: a secondary analysis of the SUMMIT trial." (2025) — PMID 39551891 — Blood volume −0.58L, systolic blood pressure −5mmHg, and CRP −37.2% at 52 weeks, post hoc exploratory correlation.
• "Tirzepatide Reduces LV Mass and Paracardiac Adipose Tissue in Obesity-Related Heart Failure: SUMMIT CMR Substudy." (2025) — PMID 39566869 — Left ventricular mass −11g versus placebo, alongside weight loss.
• "The anti-inflammatory and immunological properties of GLP-1 Receptor Agonists." (2022) — PMID 35738455 — Anti-inflammatory and immunomodulatory actions in in vitro and preclinical models, potential for new indications.
• "Efficacy and safety of oral semaglutide 14 mg (flexible dose) in early-stage symptomatic Alzheimer's disease (evoke and evoke+): two phase 3, randomised, placebo-controlled trials." (2026) — PMID 41865758 — 3,808 amyloid-confirmed MCI and mild dementia patients, CDR-SB 2.3 vs 2.3 and 2.2 vs 2.1 at 104 weeks, failed to slow progression, trial discontinued, funded by Novo Nordisk.
• "evoke and evoke+: design of two large-scale, double-blind, placebo-controlled, phase 3 studies evaluating efficacy, safety, and tolerability of semaglutide in early-stage symptomatic Alzheimer's disease." (2025) — PMID 39780249 — Design paper, primary endpoint 104-week CDR-SB change, 52-week extension.
• "Semaglutide showed limited improvements in patients with Alzheimer's disease: Revisiting the evoke and evoke + clinical trials." (2026) — PMID 42261705 — Single-investigator summary statistics reanalysis, claiming some differences at late time points such as ADCS-ADL-MCI p=0.0039 at 130 weeks, interpreted as due to blood-brain barrier penetration limits.
• "GLP-1 receptor agonism in Alzheimer's disease after EVOKE: From single hypothesis to stratified program." (2026) — PMID 42720565 — Primary and all cognitive/functional secondary endpoints failed, inference weight of extension cohort with dropouts limited, biomarker substudy interpretation complex, cardiometabolic hypothesis not refuted.
• "GLP-1 receptor agonists, metabolic syndrome, and Alzheimer's disease: Lessons and opportunities from the EVOKE trials." (2026) — PMID 42580680 — Low dementia rate in observational data, evoke failed to slow progression despite biomarker changes, early pre-symptomatic intervention studies proposed.
• "Associations of semaglutide with incidence and recurrence of alcohol use disorder in real-world population." (2024) — PMID 38806481 — 83,825 obesity EHRs, AUD incidence and recurrence 50-56% lower compared to other obesity medications (12 months).
• "Glucagon-like peptide-1 receptor agonists and risk of substance use disorders among US veterans with type 2 diabetes: cohort study." (2026) — PMID 41781010 — Veterans with type 2 diabetes, lower risk of several substance use disorders compared to SGLT-2 inhibitors, further evaluation needed.
• "Once-Weekly Semaglutide in Adults With Alcohol Use Disorder: A Randomized Clinical Trial." (2025) — PMID 39937469 — 48 patients not seeking treatment for AUD in a 9-week phase 2 trial, reduction in laboratory alcohol intake, drinks per drinking day, and craving, average daily drinks and drinking days unchanged.
• "Once-weekly semaglutide versus placebo in patients with alcohol use disorder and comorbid obesity: a randomised, double-blind, placebo-controlled trial." (2026) — PMID 42070571 — Single-center 108 patients for 26 weeks phase 2 plus cognitive behavioral therapy, heavy drinking days −41.1 vs −26.4 percentage points.
• "Oral Semaglutide for Alcohol Use Disorder: A Randomized Clinical Trial." (2026) — PMID 42522065 — 50 patients seeking treatment in an 8-week phase 2 trial, primary (laboratory craving) and daily drinks ineffective, reduction in heavy drinking days.
• "The effects of glucagon-like peptide-1 receptor agonists (GLP1-RAs) on alcohol-related outcomes: a systematic review and meta-analysis." (2025) — PMID 41350683 — 3 RCTs (430 patients) non-significant reduction in alcohol intake and craving, observational studies show reduction in alcohol-related events, larger RCTs needed.
• "Effects of glucagon-like peptide-1 receptor agonists on alcohol consumption: a systematic review and meta-analysis." (2025) — PMID 41324012 — 4 RCTs plus 10 observational studies, AUDIT reduction (high heterogeneity), large-scale dedicated RCT needed.
• "Trial of Lixisenatide in Early Parkinson's Disease." (2024) — PMID 38598572 — Phase 2 156 patients for 12 months, MDS-UPDRS III −0.04 vs +3.04, nausea 46%, longer and larger trials needed.
• "Antidiabetic drugs in Parkinson's disease: a comprehensive meta-analysis on efficacy and safety with trial sequential analysis and GRADE evaluation." (2025) — PMID 40762930 — 7 RCTs and 973 patients, no difference from placebo in MDS-UPDRS subscores, increased gastrointestinal adverse effects.
• "Semaglutide and non-arteritic anterior ischaemic optic neuropathy: Review and interpretation of reported association." (2025) — PMID 40055951 — Referral population bias leading to higher incidence in unexposed group compared to background, mixed registry networks, trial meta-analysis shows no difference, crowded optic nerve head hypothesis, causal inference not possible.
• "Association between Semaglutide and Nonarteritic Anterior Ischemic Optic Neuropathy: A Multinational Population-Based Study." (2025) — PMID 39491755 — EHRs from 160 institutions across 21 countries, no association between semaglutide and NAION.
• "Ocular complications from glucagon-like peptide-1 receptor agonists: Clinical evidence, potential mechanisms, and clinical recommendations." (2026) — PMID 41857480 — SUSTAIN-6 retinopathy worsening concentrated in pre-existing retinopathy and rapid HbA1c drop, no increase in REWIND or LEADER, EMA classifies NAION as a very rare side effect.
• "Semaglutide and Cardiovascular Outcomes in Patients with Type 2 Diabetes." (2016) — PMID 27633186 — SUSTAIN-6, 3,297 high-risk type 2 diabetes patients for 104 weeks, retinopathy complications HR 1.76.
• "Safety of Semaglutide." (2021) — PMID 34305810 — Careful monitoring for patients at risk of worsening pre-existing retinopathy due to strong blood glucose lowering (especially with insulin combination), pancreatic and thyroid cancers inconclusive.
• "Association of semaglutide with risk of suicidal ideation in a real-world cohort." (2024) — PMID 38182782 — Triggered by European regulatory agency investigation, lower risk of suicidal ideation compared to non-GLP-1 drugs in TriNetX.
• "Suicide and Self-Harm Events With GLP-1 Receptor Agonists in Adults With Diabetes or Obesity: A Systematic Review and Meta-Analysis." (2025) — PMID 40105856 — 27 RCTs, suicide and self-harm events 0.047 vs 0.042 per 100 person-years, no difference.
• "Psychiatric Safety of Semaglutide for Weight Management in People Without Known Major Psychopathology: Post Hoc Analysis of the STEP 1, 2, 3, and 5 Trials." (2024) — PMID 39226070 — Post hoc analysis of adults without major psychopathology, no increase in depression or suicidal ideation.
• "The risk of depression, anxiety, and suicidal behavior in patients with obesity on glucagon like peptide-1 receptor agonist therapy." (2024) — PMID 39424950 — Community-based cohort matched with non-use controls, reporting 195% risk increase for major depression and 106% for suicidal behavior.
• "Burden and Risk of Depression in Patients Receiving Semaglutide: A Systematic Review and Meta-Analysis." (2026) — PMID 42584177 — Depression RR 1.25 (not significant), certainty limited by high heterogeneity and bias, not proof of absence of risk.
• "Association of GLP-1 Receptor Agonists With Risk of Suicidal Ideation and Behaviour: A Systematic Review and Meta-Analysis." (2025) — PMID 39945396 — Meta-analysis of observational studies, no significant association, high heterogeneity, observation recommended for patients with psychiatric disorders.
• "Safety analysis of compounded GLP-1 receptor agonists: a pharmacovigilance study using the FDA adverse event reporting system." (2026) — PMID 40285721 — Increased suicide-related ROR 6.34, cholecystitis 3.39, compounding errors, and contamination in compounded product reports, not a causal design.
• "Glucagon-like peptide-1 receptor agonists and risk of thyroid cancer: A systematic review and meta-analysis of randomized controlled trials." (2024) — PMID 38018310 — 64 RCTs, thyroid cancer OR 1.52, fragility index 1, 5-year number needed to harm 1,349.
• "Glucagon-Like Peptide 1 Receptor Agonists and Risk of Thyroid Cancer: An International Multisite Cohort Study." (2025) — PMID 39772758 — Databases from 6 countries, no increase in thyroid cancer compared to DPP-4 inhibitors (HR 0.81), long-term risk cannot be ruled out due to 1.8-3.0 year follow-up.
• "Long-Term Glucagon-Like Peptide 1 Receptor Agonist Use Is Not Associated With Increased Risk of Thyroid Cancer in Adults With Type 2 Diabetes." (2025) — PMID 41182904 — Users of 1 year or longer with median 4.5-year follow-up, no increase in thyroid cancer.
• "Differential Effects of GLP-1 Receptor Agonists on Cancer Risk in Obesity: A Nationwide Analysis of 1.1 Million Patients." (2024) — PMID 39796706 — 1.1 million obesity patients, linked to reduced risk of multiple cancers, liraglutide thyroid cancer HR 1.70.
• "GLP-1 receptor agonists and the risk for cancer: A meta-analysis of randomized controlled trials." (2025) — PMID 40437949 — 50 RCTs, overall cancer no difference, uterine cancer reduction in obesity trials, thyroid cancer signal, short-term trial colorectal cancer increase likely due to increased screening.
• "A real-world disproportionality analysis of semaglutide: Post-marketing pharmacovigilance data." (2024) — PMID 38943656 — FAERS, unexpected signals including pancreatic cancer, intestinal obstruction, and cholecystitis.
• "Cardiovascular Effects and Tolerability of GLP-1 Receptor Agonists: A Systematic Review and Meta-Analysis of 99,599 Patients." (2025) — PMID 40892610 — 21 RCTs, reduction in death and MACE, increase in gastrointestinal and gallbladder disorders, no difference in pancreatitis or neoplasms.
• "Clinical Consequences of Delayed Gastric Emptying With GLP-1 Receptor Agonists and Tirzepatide." (2024) — PMID 39418085 — Increased gastric contents but aspiration is rare, pre-procedure management recommendations limited by lack of evidence.
• "Association between glucagon-like peptide-1 receptor agonist use and peri-operative pulmonary aspiration: a systematic review and meta-analysis." (2025) — PMID 40230298 — 28 observational studies, aspiration OR 1.04 (low certainty), gastric contents OR 5.96 (low certainty).
• "Gastrointestinal effects of GLP-1 receptor agonists: mechanisms, management, and future directions." (2024) — PMID 39096914 — Delayed gastric emptying and aspiration concerns, sparse evidence for pre-procedure guidelines.
• "Safety and Tolerability of Glucagon-Like Peptide-1 Receptor Agonists: A State-of-the-Art Narrative Review." (2026) — PMID 41351656 — NAION, pancreatitis, and medullary carcinoma associations reported mainly from pharmacovigilance and case evidence, no established increased suicide risk, limited safety data in special populations such as pregnancy, lactation, pediatrics, and severe renal/hepatic impairment.
• "Glucagon-like peptide-1 receptor agonist use in pregnancy: a review." (2025) — PMID 39181497 — Unintended pregnancy cases while taking, lack of evidence to predict the impact of exposure during pregnancy.