The word "insulin" appears in the first paragraph of every description of how GLP-1 works, and it puts a reasonable question in the mind of anyone without diabetes who is thinking about a low dose. If the drug makes the pancreas release more insulin, what stops it from dropping blood sugar too far?
The answer is built into the receptor. GLP-1 amplifies insulin release only when glucose is already elevated, and the amplification fades as glucose approaches normal. That single property separates the GLP-1 class from insulin injections and sulfonylureas, both of which lower glucose regardless of where it started, and it is why the obesity trials of semaglutide and tirzepatide, run in tens of thousands of people without diabetes at the full labelled doses, recorded clinically significant hypoglycemia in well under 1 percent of participants.
This article explains the mechanism, sets out the hypoglycemia rates from the trials, identifies the situations in which the risk becomes real, and takes apart the lightheaded, shaky feeling that people on GLP-1s often attribute to low blood sugar and that usually is not. It also looks at the opposite question, whether a GLP-1 can reduce the post meal glucose crashes known as reactive hypoglycemia. A microdose produces lower drug exposure than any of the trials cited, so the risk figures here are a ceiling rather than an estimate.
Key Takeaways
- GLP-1 receptor agonists stimulate insulin secretion in a glucose dependent way: the effect is strong when glucose is above about 140 mg/dL and close to absent below about 70 to 80 mg/dL. Insulin injections and sulfonylureas have no such gate.
- In the STEP 1, 3, and 4 trials of semaglutide 2.4 mg in adults without type 2 diabetes, clinically significant hypoglycemia (below 54 mg/dL) was reported in about 0.6 percent of participants against 0.2 percent on placebo. In SURMOUNT-1, tirzepatide produced rates below half a percent at every dose, with none on placebo.
- Nearly all hypoglycemia in the GLP-1 trials occurred in people with type 2 diabetes taking a sulfonylurea or insulin alongside the drug. In the Ozempic prescribing information, the rate on semaglutide 1 mg rose from about 3.8 percent as monotherapy to about 24 percent with a sulfonylurea and about 30 percent with basal insulin.
- Clamp studies by Nauck and colleagues in 2002 showed that GLP-1 stops stimulating insulin as glucose falls and leaves the glucagon, adrenaline, and cortisol responses to hypoglycemia intact, so the body's own defence against low glucose keeps working.
- The shaky, lightheaded feeling commonly reported on GLP-1s is more often caused by very low food intake, dehydration, or a drop in blood pressure on standing than by low glucose. A fingerstick or continuous monitor reading below 70 mg/dL at the time of symptoms is the only way to confirm true hypoglycemia.
- Fasting glucose in people without diabetes falls by about 7 to 9 mg/dL on full dose semaglutide or tirzepatide and stops at normal. GLP-1 receptor agonists slow gastric emptying, which flattens the post meal glucose peak that triggers reactive hypoglycemia, and liraglutide reduced episodes in a small trial of post bariatric hypoglycemia.
Why does GLP-1 not drive blood sugar too low?
The insulin signal is gated by glucose
GLP-1 does not drive blood sugar too low because it amplifies insulin secretion rather than triggering it, and the underlying trigger is glucose itself. When glucose enters a pancreatic beta cell it is metabolised, ATP rises, potassium channels close, the cell depolarises, calcium flows in, and insulin granules are released. GLP-1 binding to its receptor raises cyclic AMP inside the cell, which makes each of those steps more responsive. With little glucose entering the cell, there is little for cyclic AMP to amplify. The pharmacology textbooks call this a glucose dependent insulinotropic effect, and it is the property the whole class was selected for.
The threshold sits close to the normal fasting range. In hyperglycemic clamp experiments, GLP-1 infusion produces a large insulin response at 140 to 180 mg/dL, a moderate one at around 100, and a negligible one at 70 or below. Nauck and colleagues published the defining study in 2002, infusing GLP-1 or saline into healthy volunteers while stepping glucose down in stages to 45 mg/dL. Insulin secretion on GLP-1 fell away as glucose dropped through the 80s and 70s, and at the low steps the two infusions were indistinguishable. The same study measured the counter-regulatory hormones, glucagon, adrenaline, noradrenaline, cortisol, and growth hormone, and found their responses to hypoglycemia intact on GLP-1.
The glucagon result deserves emphasis because GLP-1 is known to suppress glucagon, and glucagon is the first line of defence against low glucose. The suppression turns out to be glucose dependent as well. GLP-1 lowers glucagon when glucose is high, which is useful, and stops lowering it once glucose falls below about 65 mg/dL, which is essential. Semaglutide and tirzepatide are engineered GLP-1 receptor agonists with the same receptor pharmacology, and the tirzepatide GIP component shares the glucose dependence. Nothing about a lower dose changes this; a microdose produces the same gated signal at a lower amplitude.
How Does the Insulin Response to GLP-1 Change With Glucose Level?
Illustrative insulin secretion as a percent of the maximal response, with and without GLP-1 receptor activation, across plasma glucose in mg/dL
- With GLP-1 receptor activation
- Without GLP-1
The two curves converge at the left. GLP-1 roughly doubles the insulin response at 100 to 140 mg/dL and adds almost nothing below 70, which is why the drug can lower a high glucose without pushing a normal one into hypoglycemia. The shape follows the clamp literature; the values are modelled rather than measured from any single study.
How often did hypoglycemia occur in the obesity trials?
Under 1 percent, at doses well above any microdose
Clinically significant hypoglycemia, defined as a plasma glucose below 54 mg/dL, was reported in about 0.6 percent of adults without type 2 diabetes taking semaglutide 2.4 mg across the STEP 1, 3, and 4 trials, against about 0.2 percent on placebo, according to the Wegovy prescribing information. Those trials enrolled roughly 2,600 people on the drug for 68 weeks at the highest labelled dose. The Zepbound label reports the SURMOUNT-1 figures for tirzepatide the same way: below half a percent on each of the 5, 10, and 15 mg doses, and none on placebo, across 72 weeks in about 2,500 participants. No episode in either programme was classed as severe, meaning none required assistance from another person.
The difference between 0.6 and 0.2 percent is real but small, and the labels attribute the excess to the same set of circumstances that produce low readings in anyone: a long gap without food, exercise on an empty stomach, and alcohol. What the figures rule out is a drug effect large enough to matter on its own. If GLP-1 receptor activation pushed glucose down independently of the starting level, the way insulin does, the rate at 2.4 mg semaglutide would look like the insulin add-on rate discussed below, not like placebo plus a rounding error.
These are full dose figures. A microdose protocol typically runs at a quarter to a half of the labelled maintenance dose, which lowers receptor occupancy and every downstream effect, including the already gated insulin signal. There is no trial of hypoglycemia at microdose exposures because the full dose trials found too little to study. The honest statement is that the risk at a lower dose cannot be higher than the risk at a higher dose of the same drug, and the higher dose risk is close to the population background.
What Share of Participants Without Diabetes Reported Glucose Below 54 mg/dL?
Approximate percent with clinically significant hypoglycemia in the obesity trials, drug versus placebo, from the Wegovy and Zepbound US prescribing information
- GLP-1 arm
- Placebo
The horizontal axis tops out at 1 percent. Both drugs sit at the bottom of it, at the highest doses ever tested in people without diabetes. The tirzepatide figure is the upper end of the range across the three doses.
Where does the real hypoglycemia risk come from?
Sulfonylureas, insulin, and long gaps without food
The real hypoglycemia risk on a GLP-1 comes from combining it with a medication that forces insulin regardless of glucose. In the Ozempic prescribing information, severe or blood glucose confirmed symptomatic hypoglycemia on semaglutide 1 mg was about 3.8 percent as monotherapy in people with type 2 diabetes, about 1.7 percent with metformin, about 24 percent when added to a sulfonylurea, and about 30 percent when added to basal insulin. The basal insulin figure is the revealing one: the placebo arm on the same insulin ran at about 29 percent. Insulin was producing the hypoglycemia, and semaglutide neither added much to it nor protected against it.
Sulfonylureas such as glipizide and glimepiride close the beta cell potassium channel directly, which forces insulin release whether or not glucose is elevated. A GLP-1 on top amplifies that forced release, and the combination is the highest risk pairing short of insulin. Every GLP-1 label carries the same instruction: when starting the drug in someone on a sulfonylurea or insulin, consider reducing the dose of the sulfonylurea or insulin. Metformin, SGLT2 inhibitors, DPP-4 inhibitors, and pioglitazone do not force insulin release and do not add meaningful risk. Tirzepatide follows the same pattern in the SURPASS programme, with hypoglycemia concentrated in the trial that combined it with insulin glargine.
For someone without diabetes on no other glucose lowering medication, the remaining risk factors are behavioural and would apply without the drug. Alcohol blocks hepatic glucose production for several hours and is the most common cause of hypoglycemia in people without diabetes. Prolonged fasting combined with hard exercise depletes liver glycogen. A GLP-1 changes the context because intake on the drug is often far lower than the person notices, so a drink or a training session lands on an emptier tank. The drug is not the cause, but the eating pattern it produces can set up the conditions, and that is the scenario the 0.6 percent were most likely in.
How Does Background Therapy Change the Hypoglycemia Rate in Type 2 Diabetes?
Approximate percent with severe or confirmed symptomatic hypoglycemia, semaglutide 1 mg versus placebo, by the medication it was added to, from the Ozempic US prescribing information
- Semaglutide 1 mg
- Placebo
The jump happens at the sulfonylurea and insulin bars, not at the semaglutide bars. On basal insulin the drug and placebo arms are almost level, because the insulin is setting the rate. These are type 2 diabetes populations at full labelled doses, and none of the settings apply to a person without diabetes on a GLP-1 alone.
Why do people feel shaky and lightheaded on a GLP-1 if glucose is normal?
Four causes that mimic low blood sugar
People feel shaky and lightheaded on a GLP-1 with normal glucose because several other effects of the drug produce the same adrenaline driven symptoms. The first is under eating. Appetite suppression can cut intake to 800 or 1,000 calories a day without the person registering it, and a body running on that little for days feels weak, cold, and foggy in a way that closely resembles hypoglycemia. The second is dehydration and low sodium. Fluid intake falls with food intake, nausea adds to the loss, and the result is a drop in blood volume that produces lightheadedness on standing and a racing heart. The third is orthostatic hypotension, a fall in blood pressure on standing that GLP-1s are known to produce through weight loss and mild volume depletion. The fourth is the slowed stomach itself: a meal that would normally have arrived in the blood by now has not, and the stretch and hormonal signals of an empty upper gut register as a need to eat.
Continuous glucose monitor data from people who describe these episodes typically shows glucose in the 75 to 95 mg/dL range at the time, which is normal. That observation is consistent across the CGM case reports and the clinical experience of prescribers, and it matches the trial figures: if the symptoms were hypoglycemia, the below 54 mg/dL rate in STEP would not be 0.6 percent. Whipple's triad, the clinical standard since the 1930s, requires three things for a diagnosis of hypoglycemia: symptoms, a measured low glucose at the time of the symptoms, and relief when glucose is raised. Symptoms alone do not meet it.
The distinction matters for what to do next. Someone who treats a false hypoglycemia with fast sugar gets a brief lift from the calories and a reinforced belief that the drug is dropping their glucose. Someone who measures, sees 88 mg/dL, and drinks a large glass of water with salt and eats a proper meal fixes the actual problem. A fingerstick meter costs less than a week of most protocols and settles the question in ten seconds. Two or three measured episodes with normal readings are enough to reclassify the symptom, and a reading below 70 at the time of symptoms is enough to take it seriously and raise it with a prescriber.
How Much Time Do Healthy Adults Spend Below 70 mg/dL Without Any Medication?
Approximate percent of the day in each glucose range for adults without diabetes wearing a continuous glucose monitor, from Shah et al 2019
Healthy people spend roughly 1 to 2 percent of the day below 70 mg/dL, almost all of it overnight and without symptoms. A single low reading on a continuous monitor is part of normal physiology, not evidence of a drug effect. The below 54 mg/dL slice is close to zero, which is the range the trials counted.
Can a GLP-1 reduce reactive hypoglycemia after meals?
Slower emptying removes the trigger
A GLP-1 receptor agonist can reduce reactive hypoglycemia, because the mechanism of reactive hypoglycemia is a glucose spike that arrives too fast, and slowing gastric emptying is the most direct way to blunt that spike. Reactive, or postprandial, hypoglycemia is the pattern in which glucose climbs steeply after a carbohydrate rich meal, the pancreas responds with a delayed and oversized insulin surge, and glucose falls below baseline two to three hours later. It is common after gastric bypass surgery, where food reaches the small intestine almost immediately, and it occurs in a milder form in people with early insulin resistance whose first phase insulin response is delayed.
GLP-1 receptor agonists slow gastric emptying substantially, particularly in the first weeks of treatment before partial tolerance develops. A meal that would have produced a peak of 150 mg/dL at 60 minutes instead produces a lower, later, broader curve, and the insulin response tracks it. With no overshoot, there is no undershoot. Abrahamsson and colleagues tested this in 2013 in people with severe post gastric bypass hypoglycemia, and liraglutide reduced the frequency of hypoglycemic episodes on continuous monitoring. Case series since then have reported the same with semaglutide. The evidence base is small and none of it is a labelled indication, but the direction is consistent with the pharmacology.
The relevance to a low dose protocol is that the gastric emptying effect is present at low exposures, and in fact is one of the effects that microdose protocols are often chosen for. Someone who experiences an afternoon crash after a carbohydrate heavy lunch is more likely to see that crash disappear on a GLP-1 than to see a new one appear. A continuous glucose monitor worn for two weeks before and two weeks after starting is the cleanest way to see the change in a single person, and the CGM article elsewhere on this site covers how to read the traces.
What Does Slowed Gastric Emptying Do to the Post Meal Glucose Curve?
Modelled plasma glucose in mg/dL after a 75 g carbohydrate meal in a person without diabetes, with normal emptying and with GLP-1 slowed emptying
- With GLP-1 (slowed emptying)
- Normal emptying, reactive pattern
The grey curve is a reactive pattern: a fast peak at 60 minutes and a dip toward 70 at 150 minutes. The green curve is the same meal with slowed emptying: a lower peak, a longer plateau, and no dip. The shape is modelled from published meal test and gastric emptying data rather than a single trial.
How far does fasting glucose fall in people without diabetes?
A few points, and it stops at normal
Fasting glucose in people without diabetes falls by about 7 to 9 mg/dL on full dose semaglutide or tirzepatide, and the fall stops within the normal range. In STEP 1, fasting plasma glucose dropped by about 8.4 mg/dL over 68 weeks on semaglutide 2.4 mg against about 0.5 mg/dL on placebo, from a mean baseline near 95 mg/dL. HbA1c fell by about 0.45 percentage points against 0.15 on placebo. SURMOUNT-1 reported a similar 7 to 9 mg/dL fall on tirzepatide across its three doses. In both trials the participants with prediabetes at baseline were far more likely to return to normal glucose than the placebo group, and no participant developed a persistently low fasting glucose.
The fall is a consequence of weight loss and improved insulin sensitivity more than a direct drug effect, which is why the size is modest and why it does not continue past normal. A person whose fasting glucose is 95 mg/dL and whose liver is producing glucose at a normal rate has no elevated glucose for GLP-1 to act on, so the drug's effect on fasting values is limited to whatever improvement in hepatic insulin sensitivity the weight loss brings. Someone at 108 mg/dL with prediabetes has more room to move and typically moves further. The pattern is regression toward normal, not a fixed subtraction.
This is what a fasting glucose on a microdose protocol should look like: stable or slightly lower, inside the 70 to 99 mg/dL range, with the change tracking body weight. A fasting reading that drifts below 70 on repeated mornings is not an expected result of the drug and warrants a conversation about what else is going on, most often very low intake, alcohol the evening before, or another medication.
How Much Did Fasting Glucose Change in the Obesity Trials?
Approximate mean change in fasting plasma glucose in mg/dL from baseline, participants without type 2 diabetes, STEP 1 at 68 weeks and SURMOUNT-1 at 72 weeks
A single digit fall from a normal baseline. For comparison, the same drugs lower fasting glucose by 40 to 60 mg/dL in people with type 2 diabetes starting from 170 or higher, because there the elevated glucose gives the gated mechanism something to act on.
What is worth knowing and raising with a prescriber?
Five points, none of which are a protocol
Everything below is context for a conversation rather than an instruction. The hypoglycemia rates quoted apply to the approved products at labelled doses in trial populations, and any low dose or compounded regimen needs its own assessment from the prescriber who wrote it.
A sulfonylurea or insulin changes the risk from under 1 percent to 20 or 30 percent, and the labels say to consider reducing their dose when a GLP-1 is added. Metformin does not carry this concern. The prescriber can only adjust what they know about.
A fingerstick meter or a two week continuous monitor answers whether the shaky spell is glucose or something else. Two or three episodes with readings above 70 mg/dL reclassify the symptom to under eating, dehydration, or blood pressure, each of which has a different fix.
Alcohol suppresses hepatic glucose output for hours and is the leading cause of hypoglycemia in people without diabetes. Reduced intake on a GLP-1 means the drink lands on less food, so the pattern that was harmless before may not be now.
Most of what feels like low blood sugar on a GLP-1 is an intake that has quietly fallen below 1,000 calories. A floor of protein and fluid, checked against a log, prevents the symptom rather than treating it. Protein targets are covered in the protein article on this site.
Fifteen grams of fast carbohydrate, recheck in 15 minutes, repeat if still low. A confirmed episode in someone without diabetes on a GLP-1 alone is unusual enough that the prescriber should hear about it and look for a cause beyond the drug.
What is genuinely unresolved here?
Three gaps worth naming
The trial hypoglycemia rates come from participant reports and scheduled laboratory draws, not from continuous monitoring. STEP and SURMOUNT did not put CGMs on everyone, so brief asymptomatic dips below 70 mg/dL that a monitor would have caught went uncounted in both arms. The below 54 mg/dL figures are robust because that level usually produces symptoms, but the true frequency of mild lows on the drug against placebo is not known with precision. It is likely small, because healthy people already spend 1 to 2 percent of the day there, but it has not been measured head to head.
The claim that the shaky lightheaded feeling is usually not hypoglycemia rests on CGM case reports and clinical experience rather than a systematic study. Nobody has enrolled a few hundred people on a GLP-1, given them monitors, and matched every symptomatic episode to a glucose reading. The trial rates make the claim very likely, and the alternative explanations are well documented on their own, but the direct study has not been done and the proportion of episodes with a genuinely low reading is an estimate.
Reactive hypoglycemia is the largest gap. The mechanism is sound and the post bariatric trials point the right way, but the population that experiences milder reactive dips without surgery has not been studied with a GLP-1 at any dose, and a microdose has not been studied for this purpose at all. Anyone hoping a low dose will smooth an afternoon crash is reasoning from pharmacology and a handful of small studies in a different population. That reasoning may well be right. It is not yet evidence.
Frequently Asked Questions
Selected References
- Novo Nordisk. Wegovy (semaglutide) injection, US prescribing information. Sections 5 (hypoglycemia with concomitant insulin secretagogues or insulin) and 6.1 (clinically significant hypoglycemia in STEP 1, 3, and 4).
- Novo Nordisk. Ozempic (semaglutide) injection, US prescribing information. Section 6.1, hypoglycemia by background therapy in the SUSTAIN trials.
- Eli Lilly and Company. Zepbound (tirzepatide) injection, US prescribing information. Sections 5 and 6.1 (hypoglycemia in SURMOUNT-1).
- Eli Lilly and Company. Mounjaro (tirzepatide) injection, US prescribing information. Section 6.1, hypoglycemia with insulin glargine in SURPASS-5.
- Nauck MA, Heimesaat MM, Behle K et al. Effects of glucagon-like peptide 1 on counterregulatory hormone responses, cognitive functions, and insulin secretion during hyperinsulinemic, stepped hypoglycemic clamp experiments in healthy volunteers. Journal of Clinical Endocrinology and Metabolism, 2002.
- Nauck MA, Kleine N, Ørskov C, Holst JJ, Willms B, Creutzfeldt W. Normalization of fasting hyperglycaemia by exogenous glucagon-like peptide 1 (7-36 amide) in type 2 diabetic patients. Diabetologia, 1993. Glucose dependence of the insulinotropic and glucagonostatic effects.
- Wilding JPH, Batterham RL, Calanna S et al. Once-weekly semaglutide in adults with overweight or obesity (STEP 1). New England Journal of Medicine, 2021. Fasting glucose and HbA1c secondary endpoints.
- Jastreboff AM, Aronne LJ, Ahmad NN et al. Tirzepatide once weekly for the treatment of obesity (SURMOUNT-1). New England Journal of Medicine, 2022.
- Shah VN, DuBose SN, Li Z et al. Continuous glucose monitoring profiles in healthy nondiabetic participants: a multicenter prospective study. Journal of Clinical Endocrinology and Metabolism, 2019. Time below 70 and 54 mg/dL in adults without diabetes.
- Abrahamsson N, Engström BE, Sundbom M, Karlsson FA. GLP1 analogs as treatment of postprandial hypoglycemia following gastric bypass surgery: a potential new indication? European Journal of Endocrinology, 2013.
- American Diabetes Association Professional Practice Committee. Glycemic goals and hypoglycemia: Standards of Care in Diabetes, 2024. Diabetes Care, 2024. Level 1, 2, and 3 hypoglycemia definitions and the 15-15 rule.
- International Hypoglycaemia Study Group. Glucose concentrations of less than 3.0 mmol/L (54 mg/dL) should be reported in clinical trials. Diabetes Care, 2017.
- Cryer PE. Hypoglycemia in Diabetes: Pathophysiology, Prevalence, and Prevention, 3rd edition. American Diabetes Association, 2016. Counter-regulation, alcohol, and Whipple's triad.
- Drucker DJ. Mechanisms of action and therapeutic application of glucagon-like peptide-1. Cell Metabolism, 2018. Beta cell cyclic AMP signalling and glucose dependent insulin secretion.
This article is for general educational purposes only and does not constitute medical advice, diagnosis, treatment, or a recommendation to start, stop, switch, or change the dose of any medication. Semaglutide, tirzepatide, liraglutide, and all other GLP-1 receptor agonists are prescription medicines, and the choice of product, dose, and any low dose or microdosing approach should be decided and monitored by a qualified healthcare provider, who should be told about every other medication being taken, in particular insulin and sulfonylureas. The hypoglycemia rates quoted here are taken from US prescribing information and published trials of the approved products at labelled doses in trial populations and do not apply to compounded products or to any regimen the labels do not describe. Nothing here is intended to diagnose, prevent, treat, or cure obesity, diabetes, hypoglycemia, or any other condition, and no low dose or microdosing approach described in this article has been shown in trials to be safe or effective outside the specific trial conditions cited. Several charts are illustrative models of published mechanisms rather than measured data from a single study and are labelled as such. Anyone with symptoms of low blood sugar and a measured glucose below 70 mg/dL, or with confusion, loss of consciousness, or inability to swallow, should seek medical care. Individual results vary.
Smaller Doses. Sharper Protocols.
Metabolic health you can measure, built around a dose designed to work with your body rather than against it.
Find Your Protocol