Why correct fasting stopped guaranteeing an empty stomach
Preoperative fasting rules, roughly two hours for clear liquids, six for a light meal and eight for solids or fatty food, were derived from how quickly an ordinary stomach empties. GLP-1 receptor agonists slow that process as part of how they work. A patient on semaglutide, tirzepatide or liraglutide can therefore follow the fasting instructions to the letter and still arrive in the anaesthetic room with solid food or a substantial fluid volume in the gastric antrum. When anaesthesia abolishes protective airway reflexes, that residual content is what can be regurgitated and inhaled.
That is the entire concern, and it is a narrow one. It is not a claim that these drugs are unsafe or that their metabolic benefits are in doubt. It is a claim about a specific window of a few hours during which the normal defences against inhaling stomach contents are switched off, and about whether the standard method of emptying the stomach beforehand still works in people taking a drug designed to slow the stomach down.
One thing to state before anything else: nothing in this article is an instruction. Perioperative decisions belong to the anaesthetist, surgeon or endoscopist who is actually assessing the patient in front of them, and they depend on the procedure, the urgency, the anaesthetic technique and the patient. Anyone taking one of these drugs should tell the team doing the procedure and follow what that team says, not what a reference article says.
How GLP-1 receptor agonists slow the stomach
Native GLP-1 is an incretin released from intestinal L-cells after a meal. Beyond its glucose-dependent effect on insulin secretion, it acts on receptors on vagal afferent fibres and in brainstem regions including the area postrema to coordinate a set of upper gastrointestinal responses: relaxation of the gastric fundus, reduced antral contractility and increased pyloric tone. The net result is that chyme leaves the stomach more slowly. Pharmacological agonists reproduce and exaggerate this.
The effect has been measured directly, mostly by scintigraphy and by paracetamol absorption tests that use the appearance of the drug in plasma as a proxy for delivery to the duodenum. Delay is most pronounced in the first hour or two after a meal. Short-acting agonists such as exenatide produce the largest and most durable retardation; the long-acting agents used for weight management produce a smaller but still measurable delay. Tirzepatide, which engages the GIP receptor as well, also slows gastric emptying.
There is an important nuance that is often lost. With continuous exposure to a long-acting agonist, the gastric emptying effect attenuates over weeks, a tachyphylaxis documented for liraglutide and for long-acting exenatide. The delay is therefore greatest early in treatment and after each dose escalation, and smaller in someone stable on the same dose for a year. Average values across a treated group also understate the individual variation, and for a risk that depends on the worst case rather than the mean, that tail is the part that matters.
The case reports that opened the question
Between 2022 and 2023, anaesthetists and endoscopists began publishing individual reports and small series with a consistent shape: a patient taking semaglutide, fasted according to protocol, who regurgitated solid gastric contents at induction, or who was found at upper endoscopy to have a stomach full of recognisable food many hours after their last meal. Some of these cases involved pulmonary aspiration; others involved procedures abandoned when retained food was seen.
Case reports are the weakest formal evidence there is. There is no denominator, so no rate can be calculated, and publication is heavily biased toward the alarming. Confounders travel with this population: obesity, diabetes with its own autonomic gastroparesis, opioid use and a high prevalence of reflux could each produce the same finding without the drug contributing.
What made this particular set of reports worth acting on was not their number but their coherence with known pharmacology. The compound implicated has a documented, dose-related effect on the precise physiological process behind the harm. A signal predicted by mechanism before it is observed is treated differently from one that surfaces unexplained in a spontaneous reporting database, and these reports were correctly read as a hypothesis worth testing quickly rather than as a risk estimate.
What gastric ultrasound added to the argument
Point-of-care gastric ultrasound turned anecdote into something countable. The technique images the gastric antrum with the patient supine and then in the right lateral decubitus position. A qualitative grade describes whether the antrum is empty, contains fluid only in the lateral position, or contains fluid in both; antral cross-sectional area can be converted into an estimated gastric volume. Any solid content, or an estimated volume above roughly 1.5 millilitres per kilogram, is conventionally treated as an at-risk stomach.
Applied to fasted patients arriving for elective procedures, this produced the strongest evidence in the area. Cross-sectional studies, including a widely cited 2024 analysis in a surgical journal, found residual gastric content in a substantially higher proportion of GLP-1 users than of matched non-users, with differences on the order of threefold rather than a marginal shift. These are observational studies in a population that differs systematically from non-users, so confounding by indication is live, but the effect size is large, the direction is consistent across countries, and the outcome is measured objectively from an image rather than inferred from a database code.
Why endoscopy became the sharpest test case
Upper gastrointestinal endoscopy is unusual in that it inspects the organ in question directly. Retained food is not inferred, it is seen, photographed and recorded, and the procedure is often abandoned as a result. Retrospective series comparing GLP-1 users with non-users have reported higher rates of retained gastric contents and of aborted or repeated procedures, which is a concrete, measurable cost quite separate from aspiration.
Sedation practice makes endoscopy more exposed than much of surgery. A large share of these procedures is performed under deep sedation with an unprotected airway, arguably a less defended configuration than a planned rapid sequence induction with a cuffed tracheal tube. Several large administrative database cohorts have examined aspiration pneumonia after endoscopy in GLP-1 users; the associations reported have generally been modest in relative terms and, because the baseline event rate is low, very small in absolute terms.
Endoscopy also illustrates why cancelling is not a free action. A postponed upper endoscopy can delay a cancer diagnosis, and a cancelled colonoscopy means a second bowel preparation. That asymmetry explains much of why gastroenterology bodies were more reluctant than anaesthesia bodies to endorse blanket withholding of a drug many of their patients need.
What the societies actually said, and how that changed
In mid-2023 the American Society of Anesthesiologists published rapid, explicitly consensus-based guidance, written because clinicians were asking what to do and no evidence base existed. Broadly, it advised withholding these drugs before elective procedures, with a shorter interval for daily formulations and a longer one for weekly formulations, delaying elective procedures in patients with symptoms of delayed emptying, and considering gastric ultrasound or full-stomach precautions where doubt remained. The intervals were a pragmatic judgement, not a finding.
The response from endocrinology, obesity medicine and gastroenterology was sceptical, and the objections were substantive. Withholding an incretin agonist has its own costs: deteriorating glycaemic control in people with diabetes, a disrupted titration schedule, and returning nausea when the drug is restarted. Widespread cancellations were reported, and the interval had no outcome evidence behind it, so the certainty implied by a specific number was not warranted.
In 2024 a joint document produced by anaesthesiology, bariatric surgery, gastroenterology, obesity medicine and endoscopic surgery organisations replaced that approach with risk stratification. Its general thrust is that most patients can continue their medication, that a period of clear liquids only before the procedure is a lower-cost way to reduce residual content, that individual factors such as recent dose escalation, high dose, active gastrointestinal symptoms and other causes of gastroparesis should drive the assessment, and that gastric ultrasound is useful where it is available and where the result would change the plan. Where uncertainty remains, the patient is managed as a full stomach.
Two caveats belong with any summary of this. These documents are guidance rather than regulation, they are revised as evidence accumulates, and individual hospitals adopt local protocols that may be more or less conservative. And the details are exactly what a reader should not take from an article: the specific instruction for any individual, including which dose to take and when, comes from the team performing the procedure, who know the case, the technique and their own institution's policy.
The pharmacology problem with any fixed withholding interval
Semaglutide has a terminal half-life on the order of a week, which is what makes once-weekly dosing possible. That same property undermines short withholding periods. Skipping a single weekly dose leaves the patient with roughly half the circulating drug they would otherwise have had; skipping two leaves a quarter. Reaching a genuinely negligible concentration would take something closer to a month. Tirzepatide, with a half-life of about five days, behaves similarly. Liraglutide, dosed daily with a half-life of about half a day, clears far faster, which is why daily and weekly agents are treated differently.
Nobody proposes a month off, because the harm of that is not hypothetical. Glycaemic control deteriorates, appetite and weight rebound, and restarting usually means re-climbing the titration ladder. For a patient with type 2 diabetes facing surgery, poor perioperative glucose control carries its own documented risks of infection and delayed healing.
There is also no reason to assume the gastric effect tracks plasma concentration in a simple linear way, particularly given the tachyphylaxis seen with sustained exposure. A hold long enough to matter pharmacologically may be unnecessary in a patient whose stomach was emptying normally anyway, and a hold short enough to be practical may not restore normal emptying in a patient who was markedly affected. That mismatch is the real reason the field drifted from counting days toward looking at the stomach.
How the risk gets managed on the day
The assessment starts with history. Which drug, which formulation, what dose, when was the last one, has the dose been escalated recently, and is the patient having nausea, vomiting, bloating, early satiety or a sense that food is sitting undigested. Symptoms are among the more useful discriminators available, because they identify the people in whom the gastric effect is currently pronounced rather than the people who merely hold a prescription.
The interventions available are unremarkable and long established. A preprocedural period of clear liquids only reduces the chance of solid residue. Gastric ultrasound, where trained operators exist, can triage a doubtful case into proceed, delay or take precautions. If the decision is to proceed with a possibly full stomach, the standard measures apply: rapid sequence induction, a cuffed tracheal tube rather than a supraglottic device, careful positioning, suction immediately available, and avoidance of deep sedation with an unprotected airway. Regional or neuraxial techniques sidestep the problem entirely where the operation allows.
Labels, compounded product and the disclosure problem
Semaglutide, tirzepatide and liraglutide are FDA-approved prescription medicines, with indications spanning type 2 diabetes and chronic weight management depending on the brand. Delayed gastric emptying is described in their prescribing information as a pharmacodynamic property rather than buried as an adverse event, and post-marketing experience has since added gastrointestinal motility problems including ileus to the labelling for semaglutide. Labelling in this area has continued to be updated as perioperative reports accumulated, and the current label is the authoritative version rather than any summary of it.
A complication specific to this class is that much use sits outside the ordinary prescription channel. Compounded semaglutide and tirzepatide, telehealth prescribing at a distance from any surgical team, and grey-market material sold as research chemicals all mean a patient may be taking a potent GLP-1 receptor agonist that appears on no medication list the hospital can see, sometimes at a content and concentration not reliably known even to the person taking it.
This matters more than any refinement of a withholding interval. Every part of the risk assessment above depends on the proceduralist knowing the drug is on board. A patient who does not disclose, because the product was obtained privately or because they do not think of it as a real medication, removes the assessment from the table and converts a manageable consideration into a surprise at induction. Disclosure carries no penalty; the consequence of silence can be severe.
How large is the risk in absolute terms
Pulmonary aspiration under general anaesthesia is rare. Estimates from large audits put it on the order of one event per several thousand anaesthetics in mixed surgical populations, higher in emergency and obstetric work and lower in fasted elective cases. Most events cause no lasting harm; a minority produce chemical pneumonitis, and a smaller minority lead to intensive care admission or death. Any relative increase, however real, is applied to a small base rate.
The evidence layers accordingly. That these drugs increase the chance of a non-empty stomach at a correctly fasted procedure is well supported, with large effect sizes and objective measurement. That this translates into a proportionate increase in aspiration events is plausible but far less well quantified, resting on administrative cohorts with modest associations and considerable scope for confounding. Which management strategy performs best has essentially no comparative evidence: no randomised trial has tested withholding against continuing, or ultrasound triage against a fixed interval, using clinical outcomes.
It is worth being explicit about why the missing trial is missing. With an event rate around one in several thousand and an expected relative difference unlikely to be dramatic, a randomised comparison powered on aspiration itself would need tens of thousands of patients. Guidance here is consensus filling a data gap, which is why it moved as far as it did between 2023 and 2024, and why it should be read as the best available judgement rather than settled science.