The semaglutide label reports monotherapy hypoglycemia twice, at two glucose thresholds, and the two rows disagree about which arm was worse. Over 30 weeks, documented symptomatic hypoglycemia at a threshold of 70 mg/dL occurred in 0% of 129 patients on placebo, 1.6% of 127 on 0.5 mg and 3.8% of 130 on 1 mg. Four lines further down the same table, severe or blood-glucose-confirmed symptomatic hypoglycemia at 56 mg/dL runs 1.6% on placebo and 0% on both doses.[1] Same trial, same patients, same 30 weeks. The arm that looks three times worse on one row has zero events on the other.
Which of those two rows is the one that matters
The threshold is not a stylistic choice. The American Diabetes Association and the European Association for the Study of Diabetes issued a joint position statement holding that glucose concentrations below 3.0 mmol/L — 54 mg/dL — should be reported in clinical trials.[5] A reading of 70 mg/dL is an alert value. A reading below 54 is the one the two societies converged on as clinically meaningful enough to require reporting.
By that standard the row to read is the lower one, and the lower one says the placebo arm had the events. That does not make the drug protective. With three patients across roughly 380, it makes the difference indistinguishable from nothing — which is the actual finding about hypoglycemia on a GLP-1 taken by itself. The separation between the two indications behind those trials is set out in the two-indications article.
The second molecule reports the same shape
Tirzepatide's diabetes label runs a monotherapy stratum of its own over 40 weeks, and reports at the 54 mg/dL threshold directly. Blood glucose below 54 mg/dL occurred in 1% of 115 patients on placebo and in 0% of the 121, 119 and 120 patients on 5 mg, 10 mg and 15 mg. Severe hypoglycemia was 0% in all four arms.[2] Across a threefold dose range, no dose produced a single reportable episode, and the only arm that did was the one receiving nothing.
Put the two labels side by side and the population-level picture is settled for people on no other glucose-lowering drug: at the threshold the societies specified, monotherapy produces hypoglycemia at a rate that does not separate from placebo, and in both labels the placebo arm is nominally the higher one. What moves those rates off the floor is never the injection by itself.
What changes it is a second prescription
The same tirzepatide table has a second stratum. Added to basal insulin with or without metformin over 40 weeks, glucose below 54 mg/dL occurred in 13% of 120 patients on placebo and 16%, 19% and 14% of those on 5, 10 and 15 mg.[2] The drug arms move a few points across a threefold dose range; the stratum itself moves from 0% to 13% before any tirzepatide is given.
That is the whole population-level argument in one comparison. Going from 0% to 13% is a change of background therapy. Going from 13% to 19% is a change of injection dose. The first number is more than twice the size of the second, and it is the one that was already true of the patient on the day they walked in. The management of each companion drug is worked out in the insulin article and the sulfonylurea article; what belongs here is the ranking. Hypoglycemia on this class is a finding about co-prescription, and the GLP-1 is the smaller term.
The population that buys these drugs online was not counted
Every figure above comes from trials in people with type 2 diabetes. The weight-management labels are explicit about what happened in the population without it, and the language is unusual enough to quote directly. The semaglutide weight-management label states that in its clinical trials in adult patients without type 2 diabetes for weight reduction, there was no systematic capturing or reporting of hypoglycemia.[3] The tirzepatide weight-management label says the same of its own obesity trial: there was no systematic capturing of hypoglycemia, though glucose below 54 mg/dL was recorded in 0.3% of treated patients against none on placebo.[4]
Both labels nonetheless carry a hypoglycemia warning, and both extend it by name to adults without type 2 diabetes.[3][4] The warning is real and the measurement behind it, in that population, was not made. A 0.3% figure drawn from unsystematic collection is a floor of unknown quality rather than an incidence, and it is the only number the obesity program produced.
This matters more here than it would in a clinic. Cash-pay telehealth sells overwhelmingly to people without a diabetes diagnosis, so the stratum with the best-characterized risk is the one most of these buyers are not in, and the stratum they are in is the one the trials did not measure. The contrast is direct: in the diabetes weight-reduction trial run by the same sponsor, hypoglycemia below 54 mg/dL was captured systematically and occurred in 6.2% of treated patients against 2.5% on placebo, with 10.7 episodes per 100 patient-years at 2.4 mg against 7.2 at 1 mg and 3.2 on placebo.[3] Collection was possible. It was the design of the obesity trials that did not ask for it.
Where it was measured without diabetes, and the subgroup that turns
One trial in adults without type 2 diabetes was large enough and long enough to register the event anyway. In the cardiovascular outcomes trial, 8,803 patients were exposed to semaglutide 2.4 mg for a median of 37.3 months against 8,801 on placebo for a median of 38.6 months, with safety collection limited to serious adverse events, discontinuations and events of special interest. Across that exposure, three episodes of serious hypoglycemia were reported on drug against one on placebo.[3] Three events in roughly 27,000 patient-years is the strongest available statement that, in people without diabetes and without a second glucose-lowering drug, this is a rare event.
The same paragraph of the label then names a subgroup where it is not. Among patients with a history of bariatric surgery, which the label identifies as a risk factor for hypoglycemia, serious hypoglycemia occurred in 2.3% — two of 87 — on semaglutide against 0%, none of 97, on placebo.[3] Four events in a trial of 17,604 people produce a class-level reassurance; two events in 87 produce a subgroup rate two orders of magnitude above it. Both come out of the same trial and the same label section. What happens to this class after bariatric surgery more generally is covered in the post-surgical article.
The bariatric evidence points the other way too
That subgroup is also where the literature contradicts the label most directly. A systematic review and meta-analysis of seven studies — three randomized trials, three cohort studies and one case series, covering 337 patients — examined semaglutide in post-bariatric patients with dumping syndrome or reactive hypoglycemia. It reported semaglutide reducing hypoglycemic episodes at a standardized mean difference of −0.85 (95% CI, −1.22 to −0.48, p < 0.001, I² = 35%), reducing time in hypoglycemia at −0.62 (95% CI, −0.95 to −0.29) and improving dumping symptom scores at −1.18 (95% CI, −1.64 to −0.72).[7]
So the same population carries a labeled signal of more serious hypoglycemia on drug and a pooled estimate of fewer hypoglycemic episodes on drug. The two are not strictly measuring the same thing — one counts serious adverse events in a cardiovascular trial that excluded few people, the other counts episode frequency in patients selected because they already had post-surgical reactive hypoglycemia. Neither resolves the other, and a page reporting only one of them would be describing a settled question that is not settled.
The defense is intact; the warning system is quieter
The reason monotherapy produces so little hypoglycemia has been tested directly rather than assumed. A randomized, double-blind, crossover trial put 38 adults with type 2 diabetes treated only with metformin through two 12-week periods of once-weekly semaglutide or placebo, each followed by a hypoglycemic clamp. The change in mean glucagon concentration from a plasma glucose of 5.5 mmol/L down to nadir was 88.3 pg/mL on semaglutide against 83.1 on placebo, an estimated difference of 5.2 pg/mL (95% CI, −7.7 to 18.1).[6] The body's primary counter-regulatory response to a falling glucose is not blunted by the drug.
The secondary findings in that trial are less comfortable and rarely quoted. The rise in noradrenaline and cortisol was statistically significantly lower on semaglutide than on placebo, and both the mean hypoglycemic symptom score and the proportion of participants who recognized hypoglycemia at nadir were lower on semaglutide, while cognitive function test results were similar.[6] The hormonal rescue works; the subjective alarm that tells a person to eat is measurably fainter. For anyone taking a second glucose-lowering drug, that combination is the one worth knowing, and it is why the symptoms catalogued in the dizziness article are harder to attribute on this class than they look.
What a cash intake can and cannot see
If the risk is carried by co-prescription, the intake form is the whole control. The reassuring version of the population data is that most people with diabetes know they have it. In 30,492 adults across the national survey series, confirmed undiagnosed diabetes — elevated fasting glucose and elevated HbA1c together — stood at 1.23% in 2017 to March 2020, and the share of all diabetes cases that were undiagnosed fell from 19.3% to 9.5% over three decades.[8] Roughly nine in ten cases are known to the person who has them.
The subgroup reverses that reading, and it reverses it toward exactly the people this market sells to. The same analysis found undiagnosed diabetes more prevalent in older adults, in adults with obesity, in racial and ethnic minorities and in those without health care access — and among people with diabetes, those without health care access had undiagnosed proportions ranging from 23% to 61%.[8] A buyer with obesity and no regular clinician is drawn from the part of the distribution where the 9.5% figure does not apply.
None of that is an argument against buying. It is an argument that the question determining this risk is not about the injection at all: it is whether a second glucose-lowering drug is already in the house, and whether anyone asked in words the buyer recognizes. The patterns these forms tend to miss are set out in the telehealth intake article, and the standing exclusions in who should not take these drugs.
One point is specific to the products sold here. Compounded semaglutide and tirzepatide are not FDA-approved, and the FDA does not review them for safety, efficacy or quality before a pharmacy dispenses them. Every percentage on this page was generated by approved drug at labeled doses in a monitored trial; none of it was generated by a compounded vial, and no equivalent hypoglycemia dataset exists for one. What that status does and does not mean is unpacked in the compounded comparison.
What this leaves
For an adult without diabetes taking a GLP-1 alone, the population-level evidence for hypoglycemia is thinner than the warning on the carton implies, and what exists points to a rare event: four serious episodes across 17,604 patients over three years, and monotherapy rates at the 54 mg/dL threshold that sit at zero while placebo sits at one. For an adult taking insulin or a secretagogue alongside it, the stratum jumps to 13% before the injection is added, and the injection dose barely moves it after.
The honest summary of the middle case — a person without diabetes, on no other glucose-lowering drug, buying online — is that both manufacturers put a hypoglycemia warning on a product whose obesity trials did not systematically collect the event. That is a census with its controls attached, not a reassurance and not an alarm.