Search for how a weekly GLP-1 should be injected and the answers arrive as one procedure: pick a site, rotate it, pinch the skin, go in at an angle, push slowly. Four FDA-reviewed labels cover this class, and they do not agree on four of those five points. What they disagree about is not the drug. It is the container it arrived in.
The site is free, and the labels say so plainly
Wegovy, Ozempic, Zepbound and Mounjaro are all administered under the skin of the abdomen, the thigh or the upper arm, and each label reports that similar exposure was achieved across those three regions.[6][7][8][9] Wegovy goes further and says it outright: the time of day and the injection site can be changed without the need for a dosage modification.[6] Tirzepatide’s labels carry one restriction that semaglutide’s do not, and it is about reach rather than pharmacology — the back of the upper arm is listed for another person to inject, not for self-injection.[8][9]
That single fact disposes of the most common reason people give for moving between body regions. Nobody is balancing absorption by alternating arm and thigh, because the labels record no difference to balance. If the dose itself is what you are weighing, the titration schedule is a separate question from where the needle goes.
So what is rotation for
The tirzepatide labels answer directly, and the answer is a tissue outcome. Rotate the injection site within the area chosen for each dose, they instruct, to reduce the risk of lipodystrophy — pits in the skin or thickened skin — and localized cutaneous amyloidosis, skin with lumps, at the injection sites.[8][9] The instruction is explicitly within an area: the same region of the body may be reused provided a different spot in it is chosen.
Ozempic frames the same idea from the other side, telling prescribers to instruct patients to use a different injection site each week when injecting in the same body region.[7] Wegovy is the tersest of the four, saying only to rotate injection sites with each dose and not to use the same site for each injection.[6] Novo’s labels name no lesion at all. Lilly’s name two.
The evidence behind rotation is large, and it is not about this drug
The rotation rule was built on insulin, and on that drug the observational signal is dramatic. A multicenter study examined the injection sites of 430 insulin-injecting outpatients and found lipohypertrophy in 64.4% of them. Among patients who rotated sites correctly, only 5% had it; among patients who had it, 98% either did not rotate or rotated incorrectly. Unexplained hypoglycemia was reported by 39.1% of those with lipohypertrophy against 5.9% of those without, glycemic variability by 49.1% against 6.5%, and total daily insulin averaged 56 units against 41.[3]
A later meta-analysis pooled 45 studies and 26,865 participants and put overall prevalence at 41.8% (95% CI 35.9% to 47.6%), with type 2 diabetes at 45.9% and type 1 at 39.9%. Of every moderator tested — needle reuse, needle length, injections per day, HbA1c, insulin dose, site rotation — only the duration of insulin therapy significantly influenced prevalence on meta-regression.[4] The longer the exposure, the more of it there is.
Then the randomized evidence arrives and shrinks
Teaching the technique has been tested, and the result is much smaller than the observational picture predicts. A systematic review screened 580 records and found three randomized trials, 637 participants in total, of injection-technique education in patients who already had lipohypertrophy. Education may slightly reduce the total daily dose of insulin — mean difference −6.26 units (95% CI −9.42 to −3.10, p < 0.001), rated low certainty. On HbA1c it may have little to no effect and the evidence is very uncertain: −0.59 (95% CI −1.71 to 0.54, p = 0.31, I2 = 98%). On the prevalence of hypoglycemia, little to no effect again: risk ratio 0.44 (95% CI 0.06 to 3.13, p = 0.41).[5]
Those two bodies of evidence are usually quoted as one. They are not one. The observational series separates a 39.1% hypoglycemia rate from a 5.9% rate; the randomized pool cannot distinguish a 56% reduction in hypoglycemia from a threefold increase, because its confidence interval spans both. What survives contact with randomization is a lower insulin dose. A page that cites only the first number is quoting something real and describing something false.
The census for a once-weekly GLP-1
Lipohypertrophy is a lesion of repeated insulin exposure, and insulin is injected several times a day. A weekly GLP-1 is injected 52 times a year. The indexed literature reflects that difference completely. Against 45 studies and 26,865 pooled participants on the insulin side, the records indexed under both lipohypertrophy and semaglutide number six, and all six describe semaglutide being used to treat HIV-associated lipohypertrophy — the opposite relationship to the one the rotation instruction concerns. The search instrument returns results; there is simply no published series of lipohypertrophy caused by a GLP-1 injection.
That absence is not evidence the labels are wrong to require rotation. Both manufacturers require it, one of them naming the lesions it is meant to prevent, and an instruction on an approved label is the finding here rather than a hypothesis. What the absence means is narrower: nobody has measured how much rotation matters at 52 injections a year, and any page that puts a number on it is borrowing one from a different drug at a different frequency. What the skin does after a GLP-1 injection is covered in the injection-site article, where the driver turns out not to be technique at all.
Speed is the rule that does nothing
“Inject slowly” is the most repeated instruction in this category and the one with the clearest refutation. A single-center, double-blinded randomized controlled trial gave 82 adults already receiving daily insulin or GLP-1 agonist injections 17 subcutaneous saline injections each, at 150, 300 and 450 µL per second and at volumes from 400 to 1600 µL. Injection speed had no impact on pain (p = 0.833). Volume did: 1600 against 400 µL produced a difference of 7.2 mmon a 100 mm visual analog scale (95% CI 4.6 to 9.7, p < 0.0001). Site did too, and by more: the thigh scored 9.0 mm above the abdomen (95% CI 6.7 to 11.3, p < 0.0001).[1]
A second randomized trial in 80 patients with diabetes covered a wider speed range, zero to 800 µL per second, across 24 injections each through a 27-gauge needle. Pain sensation again did not change with increasing injection speed. The thigh was 2.1 mm worse than the abdomen and the largest volume, 2250 µL, was worse than the rest, but every mean score sat below 20 mm on the 100 mm scale and acceptance ran from 93.7% to 98.7%.[2] The authors’ own reading is that the clinical impact of the differences they found is likely marginal.
Read together, the two trials say something useful and slightly deflating. Of the things a person can control at the moment of injection, speed is not one that changes the experience, and the variable that does — which site — is the one the labels leave entirely open. Weekly doses of these drugs are small in volume, which is the range in which both trials found the least pain to begin with.
Angle and pinching belong to the device
Here the four labels stop agreeing, and the pattern is worth reading closely because it is not about semaglutide or tirzepatide. The Wegovy and Ozempic single-dose prefilled syringes, which carry an integrated needle, are the only presentations in the class that state an angle: the instructions describe pinching and holding a fold of skin between thumb and fingers, holding that pinch throughout, inserting fully into the pinched skin at a 45-degree angle, and withdrawing at the same angle.[6][7]
No pen states an angle or a pinch. The Ozempic pen instructions describe wiping the site, inserting the needle and counting six seconds with the dose button held down while the counter returns to zero.[7] A Zepbound or Mounjaro single-dose pen is held against the skin with the injection button pressed for up to ten seconds; the KwikPen instruction is to push the dose knob in and slowly count to five.[8][9]
The vial-and-syringe presentation does something different again, and it is the most informative line on any of these labels. Lilly’s vial instructions say to inject exactly as the healthcare provider has shown, and that the healthcare provider should tell the patient whether to pinch the skin before injecting.[8][9] An FDA-reviewed label, asked the pinch question for the one presentation where the user controls the needle, declines to answer it and hands it to the prescriber. The needle that makes an angle matter in the first place is the subject of the needle article, and how the two presentations differ elsewhere is covered in the vial-versus-pen comparison.
All four labels do converge on a short list: not into muscle or vein, not into skin that is tender, bruised, red, hard or scaly, not into scars or stretch marks, and — on both semaglutide labels — insulin and the GLP-1 as separate injections, never mixed, acceptable in the same body region but not adjacent to each other.[6][7][8][9]
What a compounded vial changes about all of it
Everything above is quoted from labels that the FDA reviewed. A compounded preparation has none. It is not an FDA-approved drug, the agency does not review its safety, efficacy or quality before it is dispensed, and there is no reviewed insert to state an angle, a pinch, a hold time or a rotation rule.
The specific gap is sharper than that general one. A compounded GLP-1 arrives as a multi-dose vial with a separate syringe — the same presentation whose approved analog explicitly refuses to specify the pinch and refers the patient to a prescriber. On the approved side that referral lands somewhere: there is a label, a manufacturer and a dispensing pharmacist attached to a reviewed product. On the compounded side the instruction sheet in the box was written by whoever assembled the box, and what is in the vial is itself frequently undisclosed — see the contents article and, for who is making it, the vetting article.
What this settles and what it does not
It settles two claims that circulate as facts. Injection speed does not change injection pain in either randomized trial that measured it, and the injection site does not change drug exposure on any of the four labels. It also settles that there is no single GLP-1 injection technique: the angle, the pinch and the hold time are properties of the presentation, and quoting one presentation’s instruction for another is simply wrong.
It does not settle any individual injection, and nothing here is a procedure to follow. Which site suits a particular person, whether to pinch, what a specific concentration means for the volume drawn, and what to do about a site that has changed texture are questions for the prescriber who wrote the prescription and the pharmacist who filled it, both of whom can see the actual product. Where these sellers are assessed and on what criteria is set out in the methodology, and the boards themselves start with compounded semaglutide.